1 xxxx-xx-xx Oracle, Inc.
4 2.6.16.60-0.42.8 (SLES 10),
5 2.6.27.39-0.3.1 (SLES11),
6 2.6.18-194.3.1.el5 (RHEL 5)
7 2.6.18-194.3.1.0.1.el5 (OEL 5)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.30 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.10-sun2
12 * The async journal commit feature (bug 19128) and the cancel
13 lock before replay feature (bug 16774) are disabled by default.
17 Description: Handle unsent requests with rq_net_err in ptlrpc_check_set()
21 Description: apps stuck in ptlrpc_check_set() during direct I/O
22 Details : start async bulk unregistering at the same time at reply unlink
24 -------------------------------------------------------------------------------
26 2010-10-29 Oracle, Inc.
28 * Support for kernels:
29 2.6.16.60-0.69.1 (SLES 10),
30 2.6.32.19-0.2.1 (SLES11),
31 2.6.18-194.17.1.el5 (RHEL 5)
32 2.6.18-194.17.1.0.1.el5 (OEL 5)
33 * Client support for unpatched kernels:
34 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
35 2.6.16 - 2.6.32 vanilla (kernel.org)
36 * Recommended e2fsprogs version: 1.41.10-sun2
37 * The async journal commit feature (bug 19128) and the cancel
38 lock before replay feature (bug 16774) are disabled by default.
42 Description: atime is not properly updated on an MDS
44 Severity : enhancement
46 Description: Update to RHEL5.5 kernel 2.6.19-194.17.1.el5.
47 Update to OEL5.5 kernel 2.6.19-194.17.1.0.1.el5.
49 Severity : enhancement
51 Description: Update to SLES10 SP3 kernel 2.6.16.60-0.69.1.
54 Frequency : only with SLES10
56 Description: Use OFED "KMP" provided by Novell
57 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
58 Lustre is now built against this package. That means you need to
59 install the ofed-kmp package from Novell for the patchless client
60 and from our download site for the server. Note that the ofed-kmp
61 that Novell ships may not exactly match the kernel version but
62 should still be compatible.
64 Severity : enhancement
66 Description: Update SLES11 SP1 kernel to 2.6.32.19-0.2.1.
70 Description: Enabling quotas fails with non-consecutive OST numbering.
74 Description: Fix kernel warning due to lookup_on_len() called without i_mutex
79 Description: Account direct i/o inflight rpcs separately from non-direct i/o so
80 that direct i/o, which is limited by max_rpcs_in_flight, should not
81 block non-direct i/o, which is not limited by max_rpcs_in_flight.
85 Description: Fix per-NID reporting on outstanding writes
89 Description: Reduce stack pressure by uninlining some mds and ptlrpc functions.
93 Description: Remove LASSERT in lprocfs_rd_conn_uuid() since conn == NULL is a
98 Description: fix obdo leak issue in ll_setattr_raw()
102 Description: limit MMP interval
104 Severity : enhancement
106 Description: add several lfs ost enhancements
110 Description: Too many default ACLs break directory access on new directories
114 Description: Lustre inode size is not coherent across nodes.
115 Details : Update lvbo from disk when AST fails with EINVAL. Lvbo is updated
116 on EINVAL error in ldlm_handle_ast_error(). The updates in
117 filter_intent_policy() and ldlm_cb_interpret() have been removed as
122 Description: Oops at __percpu_counter_add+0x1b
123 Details : Use bdi_init()/bdi_destroy() to proper initialize backing_dev_info
128 Description: add mount option to generate 32bit ino, this can be used for 32bit
129 application compatibility.
133 Description: keep reference count for "lli_sai" to prevent it to be released
134 when "statahead_enter()"
138 Description: allow quotacheck over OSTs with sparse indices
142 Description: Objects not getting deleted for files which have been removed
143 Details : ll_have_md_lock() should differentiate between CR and CW OPEN
148 Description: pin object's inode in memory to avoid certain timeouts
152 Description: fix LBUG when obdfilter-survey is interrupted.
154 -------------------------------------------------------------------------------
156 2010-07-31 Oracle, Inc.
158 * Support for kernels:
159 2.6.16.60-0.42.8 (SLES 10),
160 2.6.27.39-0.3.1 (SLES11),
161 2.6.18-194.3.1.el5 (RHEL 5)
162 2.6.18-194.3.1.0.1.el5 (OEL 5)
163 * Client support for unpatched kernels:
164 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
165 2.6.16 - 2.6.30 vanilla (kernel.org)
166 * Recommended e2fsprogs version: 1.41.10-sun2
167 * The async journal commit feature (bug 19128) and the cancel
168 lock before replay feature (bug 16774) are disabled by default.
170 Severity : enhancement
172 Description: Update RHEL5.5 kernel to 2.6.18-194.8.1.el5 and OEL5.5 kernel to
173 2.6.18-194.8.1.0.1.el5.
175 Severity : enhancement
177 Description: using inkernel OFED stack for rhel5 & oel5.
179 Severity : enhancement
181 Description: Add "lfs_migrate" script from manual into lustre/scripts and RPMs
182 Details : lfs_migrate does a "poor man's" migration of files from their
183 current OST layout to a new OST layout as chosen by the MDS.
187 Description: mds_orphan_add_link()) error linking orphan to PENDING
188 Details : quota limits might disallow linking orphans to PENDING
189 when unlinking a file - temporary raise threads' privileges
190 when processing unlinks.
192 Severity : enhancement
194 Description: add conf-param -d option to remove permanent settings.
195 Details : Add the ability to remove permanent lctl conf_param settings.
196 (Previously conf_param settings could only be changed, not
197 removed.) This also provides a method to change failover
198 nid locations. Improve lctl man page.
200 Severity : enhancement
202 Description: add list_param to b1_8 and add "-R" option to list params
205 Severity : enhancement
207 Description: lfs quota output is not very convenient for awk/sed-parsing
208 Details : Some positions in lfs quota output table could be empty or
209 non-empty which made it hard to parse it with scripts, now a dash
210 is put instead of space where there is not supposed to be any data.
212 Severity : enhancement
214 Description: fix obdfilter-survey script to work properly with remote oss-s
216 Severity : enhancement
218 Description: add new OBDFILTER_SURVEY test suite
220 Severity : enhancement
222 Description: add new multiple mount protection (MMP) test suite
224 Severity : enhancement
226 Description: add support for async journal commit in echo client
228 Severity : enhancement
230 Description: allow userland programs to include <lustre/lustre_idl.h>
231 from stardard include directories
233 Severity : enhancement
235 Description: The prune-icache-use-trylock is no longer needed now that
236 the patch from bug 20008 is landed.
240 Description: The shrink grant feature is still active on the client although the
241 connect flag is not set.
245 Description: Don't leak grant space if the write failed with quota exceeded.
249 Description: Don't consume grant space twice on recoverable resent.
253 Description: a race condition could lead to SIGBUS being sent to an
254 application using mmap-ped files from Lustre
255 Details : truncate_complete_page implementation for the patchless
256 client could arbitrarily unset PG_Uptodate flag for a
257 page being kicked from the page cache, an uptodate check
258 right after a readpage call in filemap_fault could fail
259 because of that as though the page read had been unsuccessful.
263 Description: dlm lock slab shrinking is not efficient
264 Details : The dlm_locks slab can grow significantly and consumes a lot of
265 memory on the server. Set a hardlimit to grant_plan.
269 Description: Lustre does not do 1MB IOs to HW RAID
270 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
271 This is what we do already for SLES kernels.
275 Description: bump maximum number of phys/hw segments in the SLES11 kernel
276 until s/g chaining works properly.
281 Description: LSI Fusion MPT driver hacks to improve performance
282 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
284 Severity : enhancement
286 Description: increase default md stripe_cache_size to 16k
289 Bugzilla : 15587/21439
290 Description: don't handle security.capability xattr
291 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
292 This results in additional getxattr calls, causing VBR test
293 failures as well as a preformance drop when writing.
297 Description: obdfilter-survey is no longer working
298 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
299 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
303 Description: ll_shrink_cache does not handle __GFP_FS properly
307 Description: lfs getstripe shows wrong info for directories
308 Details : Set correct LOVEA default values for filesystem-wide.
312 Description: FSX checksum false positves due to mmap IO
313 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
314 checksum errors, if the flag is set on a request.
318 Description: file operations after eviction have successful return values
319 Details : use vfs ->flush callback to return any pending async errors
324 Description: mdsrate fails to write after 1.3+M files opened
325 Details : decrease memory usage on clients by recycling dentries and
330 Description: obdfilter-survey gives unreasonably high numbers
331 Details : Wait for all threads to complete when running test_brw.
335 Description: do not set lustre read_only device when server umount and keep
336 client records for recoverable ones
340 Description: move sync_on_lock_cancel tunable to the obdfilter layer
341 Details : move the tunable to trigger a journal flush on lock cancel from
342 the ost layer to the obdfilter layer. This tunable is useful
343 when using the async journal commit feature.
347 Description: exp->exp_nid_stats == NULL in filter_tally()
348 Details : fix race with per-nid stats by delaying procfs cleanup until
353 Description: extent lock cancellation on client can keep the cpu busy for too
358 Description: Do not fail OST activation when a llog is not found, just
359 issue an error message.
363 Description: Don't enable extents by default for MDT.
367 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
368 the AT code can access it concurrently while sending early
373 Description: Disable lockless truncate by default since it is sometimes flawed
374 and causes the write_disjoint test to fail.
378 Description: OSSs which don't have the patch from bug 20278 can trigger an
381 Severity : enhancement
383 Description: don't print message to the console when we have not managed to
388 Description: The MDS fails to synchronize OSTs which registered with the MGS
389 after the MDT. The problem is that OBD_NOTIFY_CREATE events are
390 raised too early and thus discarded by the MDT stack.
391 The fix consists of issuing OBD_NOTIFY_CREATE event in the lov
396 Description: Fix race when the ping evictor and a service thread execute
397 target_recovery_check_and_stop() concurrently.
401 Description: quota broadcast can trigger a LBUG on the MDT if there are
406 Description: Resetting the lov_objid values to last_id reported by the OST
407 during orphan recovery is incorrect and can cause the same
408 objects to be allocated twice.
410 Severity : enhancement
412 Description: "weak-modules" support
413 Details : Implement "weak-modules" support which enables kernel modules
414 to be used with any kernel that implements the same kABI. In
415 order to achieve this modules are now installed in
416 /lib/modules/$(uname -r)/updates/kernel on all distributions.
418 Severity : enhancement
420 Description: add writeconf as mount option
422 Severity : enhancement
424 Description: produce debuginfo packages for SLES.
426 Severity : enhancement
428 Description: add failover nidlist to the import proc file.
430 Severity : enhancement
432 Description: fix LUSTRE_SEQ_MAX_WIDTH for interoperability between 1.8
433 clients and 2.0 servers.
435 Severity : enhancement
437 Description: lfs find -s does not work correctly because of a bug in
442 Description: ll_read_ahead_page() must validate the dlm lock before using
447 Description: Prevent failover nids from registering with MGS first.
451 Description: fix lock inversion in ll_setattr_raw().
455 Description: object allocation is not balanced across OSTs.
456 Details : osc_precreate() should return 0, if there are enough objects left.
458 -------------------------------------------------------------------------------
460 2010-04-30 Oracle, Inc.
462 * Support for kernels:
463 2.6.16.60-0.42.8 (SLES 10),
464 2.6.27.39-0.3.1 (SLES11),
465 2.6.18-164.11.1.el5 (RHEL 5)
466 2.6.18-164.11.1.0.1.el5 (OEL 5)
467 * Client support for unpatched kernels:
468 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
469 2.6.16 - 2.6.30 vanilla (kernel.org)
470 * Recommended e2fsprogs version: 1.41.10-sun2
471 * The async journal commit feature (bug 19128) and the cancel
472 lock before replay feature (bug 16774) are disabled by default.
476 Description: open-unlinked directories trigger MDS LBUG
477 Details : Fix regression introduced by the patch from bug 19640.
478 ext3_inc_count() can reset nlink to 1 when the directory
479 is indexed and inode->i_nlink == 2. Work around the problem
480 by incrementing nlink by 2 instead of 1.
484 Description: Reconnects are not throttled
485 Details : Don't wake up pinger on reconnect failures and rely on regular
486 pings to trigger the next reconnection. Please note that the
487 pinger already uses a smaller interval if the import is
491 Frequency : only with NFS export
493 Description: Console flooded with error message from ll_inode_from_lock()
495 Details : in mds_open, initialize the child_res_id before enqueuing
496 the OPEN lock for the child inode, then to avoid senting
497 wrong ldlm_res_id to client.
501 Description: allow multiple instances of the same nid in NID hash
502 Details : Case of multiple separate clients from the same NID (as
503 with liblustre) is legitimate and so we should allow
504 multiple instances of the same NID in nid hash.
508 Description: the readahead code can sleep on a semaphore while holding a
510 Details : in ras_update, "lov_get_info" could be called during increasing
511 readahead windows, which tries to get the mutex lock "lov_lock"
512 while holding the spin_lock "ras_lock", then causes system
517 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
518 Details : fix assertion failure in the grant code.
522 Description: Use CNETERR (which is rate limited) in specific places in
523 the portal's LNET driver to avoid flooding the console.
527 Description: include last created object in precreate slow case
531 Description: don't do rep-ack if not created anything
532 Details : mds_open currently always put a lock into a rep-ack regardless
533 if something was created or not. This is pointless and only
534 creates needless contention. In fact the entire idea was to do
535 this for real creates as a recovery protection.
539 Description: Spurious error messages from smp_processor_id() on preemptible
541 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
542 first. The function fs_trace_get_tcd() was moved up.
546 Description: interval_erase() fix
547 Details : interval_erase() calls update_maxhigh() properly when child
552 Description: "lfs df" does not print stats for all mountpoints
553 Details : Print all mounted lustre filesystems with "lfs df"
557 Description: lfs setstripe -p not longer work with a relative pathname
558 Details : Use realpath() to provide absolute pathname.
562 Description: fix for truncated reply buffer
563 Details : reply buffer could be referred by reply_in_callback after released
568 Description: lustre.lov error when backing up symlinks with extended attributes
569 Details : Improved logic in ll_listxattr()
573 Description: properly handle null value for setattr -n lustre.lov
574 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
575 in ll_setxattr() due to no checking if "value" is NULL.
576 This command now resets to the default striping when executed
581 Description: stack overflow on lock cancellation due to fsync call
582 Details : sync_on_lock_cancel is needed for recovery when async journal
583 is enabled, but we actually just need to make sure that
584 metadata blocks have hit the journal, so doing a fs sync
585 should be enough and should consume less stack (just create an
586 empty handle and commmit it).
590 Description: using current->journal_info to store per-thread data leads
591 to problem under memory pressure
592 Details : disable the per-thread data (current->journal_info) containing
593 the lock info during I/O to work around the issue for short tem
597 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
598 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
599 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
604 Description: Cannot send after transport shutdown
605 Details : Clear imp_vbr_failed flag upon eviction
609 Description: soft lock in request set code during recovery
610 Details : during recovery, uses req->rq_set itself to replay the request
611 instead of ptlrpcd_recovery_pc
615 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
616 Details : During an ll_readahead under ll_readpage, we have seen the the
617 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
618 a page that is already locked by the readahead code.
623 Description: stop waitting for next replay transno if shutdown
624 Details : if the system is shutting down, wake up service thread blocked
625 to wait for next replay transno during recovery, then all the
626 references held by queued requests can be dropped and device
631 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
632 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
636 Description: per-nid stats should not access lustre-hash internal structures
641 Description: mount.lustre fails to pass some options to mount()
646 Description: ext4 extent allocation is slower than in ext3
647 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
648 enlarge ext4 preallocation table for 2048 4K blocks extents
653 Description: incorrect triggering of synchronous IO
654 Details : The OSC can mistakenly fall back to synchronous IO when the
655 max_dirty_mb limit is reached and no write requests have yet
656 been issued. This can occur when the dirty pages are spread
657 over many files all of which are below the optimal request size.
661 Description: Optimize quota_ctl operations by sending requests in parallel
662 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
663 Compiled from two attachments in the ticket.
667 Description: ordering issue between transaction start & i_mutex
668 Details : start the transaction earlier in llog_lvfs_destroy to get
669 transaction start and inode mutex lock nested properly.
673 Description: lru resize SLV can get stuck
674 Details : calculate SLV with a greater precision to not lose small
675 changes due to interger math truncation; round up SLV only
676 if the amount of granted locks less than the limit to not
677 get stuck with this SLV
681 Description: avoid divide-by-zero in lprocfs_rd_import()
685 Description: lfs quota failed when OSTs are down
686 Details : really return approximate block/inode usage when OSTs are down
690 Description: llobdstat fix and enhancement
691 Details : add a counter to set a limit to how many samples will be returned
692 fix a wildcard in the path to limit to obdfilter stats only
696 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
698 Details : abort lock enqueue processing sooner when we umount is in progress.
700 Severity : enhancement
702 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
703 2.6.18-164.11.1.0.1.el5.
705 Severity : enhancement
707 Description: error message improvements
708 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
709 Simplify MDT/OST service start message.
710 Suppress "changing the import ..." warning.
713 Bugzilla : 21961/17914
714 Description: ignore trailing -mdc when determining index number
717 Frequency : only with SLES11
719 Description: fix for a race condition in linux quotas implementation
720 Details : dq_flags(struct dquot) access is not properly locked which could
721 lead to certain inconsistencies when accessing it using non-atomic
722 bit operations like __set_bit in do_set_dqblk.
726 Description: return any pending async errors in close(2) using flush callback
728 -------------------------------------------------------------------------------
730 2010-01-29 Sun Microsystems, Inc.
732 * Support for kernels:
733 2.6.16.60-0.42.8 (SLES 10),
734 2.6.27.39-0.3.1 (SLES11),
735 2.6.18-164.11.1.el5 (RHEL 5)
736 2.6.18-164.6.1.0.1.el5 (OEL 5)
737 * Client support for unpatched kernels:
738 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
739 2.6.16 - 2.6.30 vanilla (kernel.org)
740 * Recommended e2fsprogs version: 1.41.6.sun1
741 * The async journal commit feature (bug 19128) and the cancel
742 lock before replay feature (bug 16774) are disabled by default.
744 Severity : enhancement
746 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
747 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
749 Severity : enhancement
750 Bugzilla : 21511/19848
751 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
753 Severity : enhancement
755 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
757 Severity : enhancement
759 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
760 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
762 Severity : enhancement
764 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
767 Severity : enhancement
769 Description: Vanilla kernel 2.6.30 patchless client support.
774 Description: bad entry in directory xxx: inode out of bounds
775 Details : fix locking issue in the rename path which could race with any
776 other operations updating the same directory.
778 Severity : enhancement
780 Description: Make watchdog timer messages to be more clear and descriptive.
784 Description: cp -p command does not preserve the dates and timestamp
785 Details : mtime could be spoiled by a write callback
789 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
793 Description: Allow non-root access for "lfs check".
794 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
796 Severity : enhancement
798 Description: quotacheck performance/scaling issues
799 Details : reduce quotacheck time on empty filesystem by skipping uninit
802 Severity : enhancement
804 Description: Enhancement for lfs(1) command to use numeric uid/gid.
806 Severity : enhancement
808 Description: Adjust locks' extents on their first enqueue, so that at the time
809 they get granted, there is no need for another pass through the
810 queues since they are already shaped into the proper forms.
814 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
815 arguments and prevent LBUG() in lustre_shrink_reply_v2().
819 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
820 exactly the specified mount options are used. Leaving off
821 any "mandatory" mount options is an error. Leaving off any
822 default mount options causes a warning, but is allowed.
823 Change errors=remount-ro from mandatory to default. Sanitize
824 the mount string before storing it. Update man pages accordingly.
828 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
829 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
832 Severity : enhancement
834 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
836 Severity : enhancement
838 Description: Clients can replay thousands of unused locks during recovery
839 Details : Don't replay unused locks (only read locks for now) during
840 recovery. This feature is disabled by default and can be
841 enabled by running the following command on the clients:
842 lctl get_param ldlm.cancel_unused_locks_before_replay
846 Description: can't stat file in some situation.
847 Details : improve initialize osc date when target is added to mds and
848 ability to resend too big getattr request is client isn't have info
853 Description: Prevent inconsistences between linux and lustre mount structures.
854 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
855 Make the sleep interruptible.
857 Severity : enhancement
859 Description: Communicate OST degraded/readonly state via statfs to MDS
860 Details : Flags in the statfs returned from OSTs indicate whether the
861 OST is in a degraded RAID state, or if the filesystem has
862 turned read-only after a filesystem error is detected.
867 Description: don't panic if EPROTO was hit when reading symlink
868 Details : correctly handling request reference in error cases.
873 Description: open sometimes returns ENOENT instead of EACCES
874 Details : checking permission should be part of open part of mds_open, not
875 lookup part. so server should be set DISP_OPEN_OPEN disposition
876 before starting permission check.
877 Also not need revalidate dentry if client already have LOOKUP lock.
881 Description: enable client interface failover
882 Frequency : on servers with multiple network interfaces
883 Details : When a child reconnects from another NID, properly update export
884 nid hash position and ldlm reverse import.
886 Severity : enhancement
888 Description: implemented direct I/O with arbitrary (nonaligned) memory
889 addresses and file offsets.
891 Severity : enhancement
893 Description: added more recovery timeout options.
895 Severity : enhancement
897 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
902 Description: Avoid deadlock for local client writes
903 Frequency : only on systems with clients writing to an OST on the same node
904 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
905 memory freeing context. This allows OST threads to set the
906 PF_MEMALLOC flag on task structures in order to allocate memory
907 from reserved pools and complete IO.
908 Use GFP_HIGHUSER for OST allocations for non-local client writes,
909 so that the OST threads generate memory pressure and allow
910 inactive pages to be reclaimed.
915 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
916 Details : move ldlm namespace creation in setup phase to avoid grab
917 _lprocfs_lock with cli_sem held.
921 Frequency : only during format of test systems
922 Description: Unable to run several mkfs.lustre on loop devices at the same time
923 Details : mkfs.lustre returns error 256 on the concurrent loop devices
924 formatting. The solution is to proper handle the error.
926 Severity : enhancement
928 Description: implement async create (obd_async_create) method for osc, to avoid
929 too long waiting new ost objects with holding ldlm lock.
933 Frequency : occasionally during network problems
934 Description: client not allowed to reconnect to OST because of active request
935 Details : abort bulk requests received by the OST once the client has timed
936 out since the client will resend the request anyway. The client
937 also now retries to reconnect to the same server if a connect
938 request failed with EBUSY or -EAGAIN.
941 Frequency : rare, if used wide striped file and one ost in down.
943 Descriptoin: don't return error if we created a subset of objects for file.
944 Details : lov_update_create_set() uses set->set_success as index for created
945 objects, so if some requests failed, they will have hole at end of
946 array and we can use qos_shrink_lsm for allocate correct lsm.
950 Description: Slow stale export processing during normal start up
951 Details : The global mgc lock prevents OST setup to be run in parallel.
952 Replace the global lock with a per-config_llog_data semaphore.
956 Description: Out or order replies might be lost on replay
957 Details : In ptlrpc_retain_replayable_request if we cannot find retained
958 request with tid smaller then one currently being added, add it
959 to the start, not end of the list.
963 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
964 Details : add cond_resched() calls to avoid hogging the cpu for too long
965 in the hash code. Make also lustre_hash_for_each_empty() more
968 Severity : enhancement
970 Description: Performance improvements for debug messages with D_RPCTRACE,
971 D_LDLM, D_QUOTA options.
974 Frequency : only with NFS export
976 Description: (lov_merge.c:74:lov_merge_lvb())
977 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
978 Details : Fix a race in the nfs export code by populating inode
979 info while the new inode is still locked
981 Severity : enhancement
983 Description: add a new file in procfs called force_lbug. Writting to this
984 file triggers a LBUG. Only for test purpose.
988 Description: OOM killer causes node hang
989 Details : really interrupt the sleep in osc_enter_cache on signals
993 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
994 Details : fix race during quota release on the slave.
996 Severity : enhancement
998 Description: smaller hash bucket sizes, cleanups
999 Details : increase hash table sizes and enabled rehashing for pools, quota,
1000 uuid, nid & per-nid stats.
1002 Severity : enhancement
1004 Description: Add ldiskfs maxdirsize mount option
1005 Details : add max_dir size mount option
1009 Description: panic in ll_statahead_thread
1010 Details : prevent parent thread to be killed before its child
1013 Frequency : only with 16TB device
1015 Description: unable to perform "mount -t lustre" of 16TB OST device
1016 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
1020 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
1021 Details : unregistering should be zero if no RPC inflight.
1025 Description: hyperion: Oops during metabench
1026 Details : Correct the refcount of lov_request_set
1028 Severity : enhancement
1030 Description: Add mptlinux and nxge drivers to Lustre builds
1032 Severity : enhancement
1034 Description: Fix watchdog timer message to be more clear
1035 Details : Make watchdog timer messages more clear and descriptive.
1039 Description: LNET soft lockups in socknal_cd thread
1040 Details : don't hog CPU for active-connecting if another connd is
1041 accepting connecting-requst from the same peer
1045 Description: recovery-small test_17 hang
1046 Details : Land several AT improvements & fixes.
1050 Description: MDS panic and hanging client processes
1051 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
1055 Description: OSS stuck in recovery.
1056 Details : fix race during recovery. class_unlink_export,
1057 class_set_export_delayed and target_queue_last_replay_reply
1058 may race while increasing/decreasing obd_recoverable_clients
1059 and obd_delayed_clients, causing recovery to wait forever.
1061 Severity : enhancement
1063 Description: add cascading_rw.c to lustre/tests
1067 Description: filter_last_id() NULL deref
1068 Details : lprocfs_filter_rd_last_id() should check for the fully
1069 setup obd device, before proceeding further.
1071 Severity : enhancement
1073 Description: Loadgen improvements
1074 Details : stacksize and locking fixes for loadgen
1078 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
1079 Details : The atomic_read() allowing the atomic_inc() are not covered
1080 by a lock. Thus they may safely race and trip this CERROR()
1081 unless we add in a small fudge factor (+1).
1083 Severity : enhancement
1085 Description: shrink_slab: nr=-9223362083340912175
1086 Details : fix spurious message from shrink_slab reporing negative nr
1090 Description: Quiet bogus previously committed transno error
1091 Details : suppress the "server went back in time" error message which
1092 is always printed even in the common case after a client eviction
1094 Severity : enhancement
1096 Description: Parallel statfs() calls result in client eviction
1097 Details : cache statfs data for 1s.
1101 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
1103 Details : fix serveral issues in pinger code causing clients not to ping
1104 servers for too long, resulting in evictions.
1108 Description: e2fsck should warn when MMP update interval is extended
1109 Details : print mmp_check_interval and make it possible to abort mount
1110 operation in case it takes too long.
1114 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
1115 Details : fix bug in the RHEL5's jbd2 callback patch.
1119 Description: drop number of active requests when queued for recovery
1120 Details : Now that we take a reference on the original request instead of
1121 making a copy of it for recovery. We need to drop the number of
1122 active requests or the queued requests will prevent all request
1123 processing when they exceed (srv->srv_threads_running - 1).
1125 Severity : enhancement
1127 Description: refuse to invalidate operational quota files when they are in use
1128 Details : an attempt to invalidate operational quota files on the quota
1129 master is not actually permitted by VFS (returning -EPERM), but we
1130 should not depend on that and should return the error earlier.
1134 Description: Applications stuck in jbd2_log_wait_commit during exit
1135 Details : fix deadlock between kjournald2 trying to acquire the page lock
1136 owned by an ost_io thread waiting for journal commit.
1138 -------------------------------------------------------------------------------
1140 2009-10-16 Sun Microsystems, Inc.
1142 * Support for kernels:
1143 2.6.16.60-0.42.4 (SLES 10),
1144 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
1145 2.6.18-128.7.1.el5 (RHEL 5),
1146 2.6.18-128.7.1.el5 (OEL 5),
1147 * Client support for unpatched kernels:
1148 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1149 2.6.16 - 2.6.27 vanilla (kernel.org)
1150 * Recommended e2fsprogs version: 1.41.6.sun1
1151 * ext4 support for RHEL5 is experimental and thus should not be
1154 Severity : enhancement
1156 Description: Add OEL5 support.
1158 Severity : enhancement
1160 Description: Update kernel to SLES11 2.6.27.29-0.1.
1164 Description: File checksum failures with OST read cache on
1165 Details : Disable page poisoning when the bulk transfer has to be aborted
1166 because the client got evicted.
1170 Description: Don't allow make backward step on assiging osc next id.
1171 Details : race between allocation next id and ll_sync thread can be cause
1172 of set wrong osc next id and can be kill valid ost objects.
1174 Severity : enhancement
1176 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1178 Severity : enhancement
1180 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
1184 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1188 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1192 Description: Error handling in osc_statfs_interpret() has been improved.
1193 Details : Check in osc_statfs_interpret() for EBADR.
1197 Description: Do not send statfs() requests to OSTs disabled by administrator.
1198 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1202 Description: Do not update ctime for the deleted inode.
1203 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
1207 Description: Increase of the size of the LDLM resource hash.
1208 Details : Bump up RES_HASH_BITS=12.
1212 Description: correctly send lsm on open replay
1213 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1218 Description: Deadlock between filter_destroy() and filter_commitrw_write().
1219 Details : filter_destroy() does not hold the DLM lock over the whole
1220 operation. If the DLM lock is dropped, filter_commitrw() can go
1221 through, causing the deadlock between page lock and i_mutex.
1222 The i_alloc_sem should also be hold in filter_destroy() while
1223 truncating the file.
1227 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1228 Details : ldiskfs_truncate calls grab_cache_page which may start page
1229 allocation under an open transaction. This may lead to
1230 calling prune_icache with consequent lustre reentrance.
1233 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
1234 still up and when the OST pool feature is used
1236 Description: interop testing got LBUG when run dd with OST pool
1237 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
1238 Details : down/upgrading the MDS to a version that doesn't/does support OST
1239 pool can cause clients to crash because the lsm has changed
1244 Description: missing tree_status on 1.8.1 RPM build
1245 Details : make rpms failed due because the tree_status file is missing.
1249 Description: continuing LustreError "mds adjust qunit failed!"
1250 Details : don't print message on the console when ->adjust_qunit fails.
1254 Description: don't increase ldlm timeout if previous client was evicted
1255 Details : if a client doesn't respond to a blocking callback within the
1256 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
1257 when the lock is next granted.
1261 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1262 affects recovery negatively.
1263 Details : make sure all exports have been properly destroyed by the zombie
1264 thread processed before stopping the target.
1268 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1269 Details : disable write barrier for ext4/SLES11.
1273 Description: Kernel panic - not syncing: Out of memory and no killable
1274 processes... on OSS when iozone
1275 Details : fix memory leak in the journal checksum patch.
1279 Description: group quota "too many blocks" OSS crashes
1280 Details : we should keep the same uid/gid for lquota_chkquota() and
1281 lquota_pending_commit()
1285 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1286 Details : don't LBUG on release quota error. Just a workaround until the
1287 problem is understood.
1289 ------------------------------------------------------------------------------
1290 2009-07-31 Sun Microsystems, Inc.
1292 * Support for kernels:
1293 2.6.16.60-0.39.3 (SLES 10),
1294 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1295 2.6.18-128.1.14.el5 (RHEL 5),
1296 * Client support for unpatched kernels:
1297 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1298 2.6.16 - 2.6.27 vanilla (kernel.org)
1299 * Recommended e2fsprogs version: 1.41.6.sun1
1300 * File join has been disabled in this release, refer to Bugzilla 16929.
1301 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1302 of Lustre filesystem with 4K stack may cause a stack overflow. For
1303 more information, please refer to bugzilla 17630.
1304 * ext4 support for RHEL5 is experimental and thus should not be
1307 Severity : enhancement
1309 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1312 Frequency : with 1.8 server and 1.6 clients
1314 Description: correctly shrink reply for avoid send too big message to client.
1315 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1316 some problems with sending this reply to 1.6 client.
1320 Description: Repeated atomic allocation failures.
1321 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1322 to generate memory pressure and allow reclaiming of inactive pages.
1323 At the same time, do not allow to exhaust emergency pools.
1324 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1326 Severity : enhancement
1327 Bugzilla : 19846, 18289
1328 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1330 Severity : enhancement
1331 Bugzilla : 19625, 16893, 18668, 19848
1332 Description: Add support for SLES11 2.6.27.23-0.1.
1334 Severity : enhancement
1336 Description: Update client support to vanila kernels up to 2.6.27.
1338 Severity : enhancement
1340 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1342 Severity : enhancement
1344 Description: Compile with -Werror by default for i686 and x86_64.
1348 Description: resolve race between obd_disconnect and class_disconnect_exports
1349 Details : if obd_disconnect will be called to already disconnected export he
1350 forget release one reference and osc module can't unloaded.
1352 Severity : enhancement
1354 Description: move AT tunable parameters for more consistent usage
1355 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1359 Descriptoin: correctly skip time estimate if in recovery
1360 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1364 Descriptoin: OSS DeadLock
1365 Details : Use trylock to prevent deadlock when shrink icache.
1367 Severity : enhancement
1369 Description: Allow tuning service thread via /proc
1370 Details : For each service a new
1371 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1372 created that can be used to set min/max thread counts, and get the
1373 current number of running threads.
1375 Severity : enhancement
1377 Description: Add state history info file, enhance import info file
1378 Details : Track import connection state changes in a new osc/mdc proc file;
1379 add overview-type data to the osc/mdc import proc file.
1383 Description: Reduce small size read RPC
1384 Details : Set read-ahead limite for every file and only do read-ahead when
1385 available read-ahead pages are bigger than 1M to avoid small size
1390 Description: free_entry erroneously used groups_free instead of put_group_info
1392 Severity : enhancement
1394 Description: Make read-ahead stripe size aligned.
1396 Severity : enhancement
1398 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1401 Frequency : rare, connect and disconnect target at same time
1403 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1404 Details : don't call obd_disconnect under lov_lock. this long time
1405 operation and can block ptlrpcd which answer to connect request.
1408 Frequency : start MDS on uncleanly shutdowned MDS device
1410 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1411 Details : stay in waiting mds<>ost recovery finished produce random bugs
1412 due race between two ll_sync thread for one lov target. send
1413 ACTIVATE event only if connect realy finished and import have
1417 Frequency : start MDS on uncleanly shutdowned MDS device
1419 Descriptoin: aborting recovery hang on MDS
1420 Details : don't throttle destroy RPCs for the MDT.
1424 Description: Slow reads beyond 8Tb offsets.
1425 Details : Page index integer overflow in ll_read_ahead_page
1429 Description: Soft lockup on OSS after MDS failover
1430 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1431 As a conseqence, the patch from bug 18224 is not operational
1432 and the MDS export cannot be reused on the OSTs until it gets
1436 Frequency : rare, only if using MMP with Linux RAID
1438 Description: MMP doesn't work with Linux RAID
1439 Details : While using HA for Lustre servers with Linux RAID, it is possible
1440 that MMP will not detect multiple mounts. To make this work we
1441 need to unplug the device queue in RAID when the MMP block is being
1442 written. Also while reading the MMP block, we should read it from
1443 disk and not the cached one.
1446 Frequency : rare, during recovery
1448 Description: Assertion failure in ldlm_lock_put
1449 Details : Do not put cancelled locks into replay list, hold references on
1450 locks in replay list
1454 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1455 Details : create_count always drops to the min value (=32) because grow_count
1456 is being changed before the precreate RPC completes.
1459 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1462 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1463 Details : add proper locking for creation of jbd_4k slab cache
1467 Description: MMP check in ext3_remount() fails without displaying any error
1468 Details : When multiple mount protection fails during remount, proper error
1473 Description: Rare Client crash on resend if the file was deleted.
1474 Details : When file is opened, but open reply is lost and file is
1475 subsequently deleted before resend, resend processing logic
1476 breaks trying to open the file again, should not try to open.
1480 Description: add check for >8TB ldiskfs filesystems
1481 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1482 Don't allow >8TB ldiskfs filesystems to be mounted without
1483 force_over_8tb mount option
1487 Description: Client locked up when running multiple instances of an app. on
1488 multiple mount points
1489 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1490 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1494 Description: Cannot acces an NFS-mounted Lustre filesystem
1495 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1496 from a Lustre-client exporting the Lustre filesystem via NFS.
1500 Description: panic in ll_statahead_thread
1501 Details : grab dentry reference in parent process.
1504 -------------------------------------------------------------------------------
1506 tbd Sun Microsystems, Inc.
1508 * Support for kernels:
1509 2.6.16.60-0.31 (SLES 10),
1510 2.6.18-128.1.6.el5 (RHEL 5),
1511 2.6.22.14 vanilla (kernel.org)
1512 * Client support for unpatched kernels:
1513 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1514 2.6.16 - 2.6.22 vanilla (kernel.org)
1515 * Recommended e2fsprogs version: 1.40.11-sun1
1516 * File join has been disabled, refer to Bugzilla 16929.
1517 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1518 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1519 Lustre filesystem with 4K stack may cause a stack overflow. For more
1520 information, please refer to bugzilla 17630.
1522 Severity : enhancement
1524 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1526 Severity : enhancement
1528 Description: Update OFED release to 1.4.1 RC4
1530 Severity : major, only with big OST
1532 Description: Very poor metadata performance on Infiniband lustre configuration
1533 Details : OST object precreation becomes very slow on big OSTs. This is due
1534 to the ialloc patch spending too much time scanning groups.
1537 Frequency : during recovery
1539 Description: don't mix llog inodes with normal.
1540 Details : allocate inodes for log in last inode group
1545 Description: fix lqs' reference which won't be put in some situations
1546 Details : This patch fixes:
1547 1. In quota_check_common(), this function will check quota
1548 for user and group, but only send one return via "pending".
1549 In most cases, the pendings should be same. But that is not
1551 2. If quotaoff runs between lquota_chkquota() and
1552 lquota_pending_commit(), the same thing will happen too.
1553 That is why it comes:
1554 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1557 -------------------------------------------------------------------------------
1559 2008-12-31 Sun Microsystems, Inc.
1561 * Support for kernels:
1562 2.6.5-7.314 (SLES 9),
1563 2.6.9-67.0.22.EL (RHEL 4),
1564 2.6.16.60-0.31 (SLES 10),
1565 2.6.18-92.1.17.el5 (RHEL 5),
1566 2.6.22.14 vanilla (kernel.org)
1567 * Client support for unpatched kernels:
1568 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1569 2.6.16 - 2.6.22 vanilla (kernel.org)
1570 * Client support for unpatched kernels:
1571 we do not recommend using patchless RHEL4 clients with kernels
1572 prior to 2.6.9-55EL (RHEL4U5).
1573 * Recommended e2fsprogs version: 1.40.11-sun1
1574 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1575 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1576 removed cwd "./" (refer to Bugzilla 14399).
1577 * A new quota file format has been introduced in 1.6.5.
1578 The format conversion from prior releases is handled transparently,
1579 but releases older than 1.4.12/1.6.5 don't understand this new
1580 format. The automatic format conversion can be avoided by running
1581 the following command on the MDS:
1582 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1583 For more information, please refer to bugzilla 13904.
1584 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1585 The format conversion from prior releases is handled transparently,
1586 but releases older than 1.6.6/1.8.0 don't understand this new
1587 format. The automatic format conversion can be avoided by running
1588 the following commands on the MDS and OSS servers (for
1589 pre 1.4.12-1.6.5 quota files):
1590 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1591 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1592 or (for 1.4.12/1.6.5 quota files)
1593 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1594 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1595 For more information, please refer to bugzilla 13904.
1596 * Output of lfs quota has been made less detailed by default,
1597 old (verbose) output can be obtained by using -v option.
1598 * File join has been disabled in this release, refer to Bugzilla 16929.
1599 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1600 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1601 Lustre filesystem with 4K stack may cause a stack overflow. For more
1602 information, please refer to bugzilla 17630.
1604 Severity : enhancement
1606 Description: Caching OSS
1607 Details : introduce data caching on the OSS. The OSS now relies on the linux
1608 kernel page cache to keep recently accessed data in memory.
1609 It is worth noting that all write requests are still flushed
1610 synchronously as in lustre 1.6.
1612 Severity : enhancement
1614 Description: Version based recovery
1615 Details : introduce finer grained recovery able to detect transaction
1616 dependencies and can deal with transaction gaps caused by clients
1617 failing at the same time as the server.
1619 Severity : enhancement
1621 Description: Enable adaptive timeouts by default
1622 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1623 managed dynamically based on server load and should not need
1624 to be tuned manually based on cluster size. This allows Lustre
1625 to work under a wider variety of system sizes and loads, without
1626 unnecessarily causing lengthy recovery times.
1628 Severity : enhancement
1630 Description: Add OST Pools support
1631 Details : File striping can now be set to use an arbitrary pool of OSTs
1633 Severity : enhancement
1635 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1636 Details : allow skip disconnected ost for send statfs request and hide error
1640 Frequency : rare, on llog test 6
1642 Descriptoin: don't allow connect to already connected import
1643 Details : allowing connect to already connected import is hide connecting problem.
1646 Frequency : rare, on failed llog setup
1648 Descriptoin: don't leak obd reference on failed llog setup
1649 Details : for failed llog setup - mgc forget call class_destroy_import for
1650 client import, move destroy import to more generic place.
1655 Descriptoin: allow kill process which wait statahead result
1656 Details : for some reasons 'ls' can stick in waiting result from statahead,
1657 in this case need way for kill this process.
1660 Frequency : rare, at shutdown
1662 Descriptoin: panic at umount
1663 Details : llap_shrinker can be raced with killing super block from list and
1664 this produce panic with access to already freeded pointer
1669 Descriptoin: don't lose wakeup for imp_recovery_waitq
1670 Details : recover_import_no_retry or invalidate_import and import_close can
1671 both sleep on imp_recovery_waitq, but we was send only one wakeup
1677 Descriptoin: panic in mds_open
1678 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1683 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1685 Details : release lock refernce only after add page to pages list.
1688 Frequency : always with long access acl
1690 Descriptoin: mds can't pack reply with long acl.
1691 Details : mds don't control size of acl but they limited by reint/getattr
1696 Frequency : on remount
1698 Description: external journal device not working after the remount
1699 Details : clear dev_rdonly flag for external journal devices in
1705 Description: shutdown vs evict race
1706 Details : client_disconnect_export vs connect request race.
1707 if client will evicted at this time - we start invalidate
1708 thread without referece to import and import can be freed
1714 Description: shrink LOV EAs before replying
1715 Details : correctly adjust LOV EA buffer for reply.
1720 Description: don't skip ost target if they assigned to file
1721 Details : Drop slow OSCs if we can, but not for requested start idx.
1722 This means "if OSC is slow and it is not the requested
1723 start OST, then it can be skipped, otherwise skip it only
1724 if it is inactive/recovering/out-of-space.
1726 Severity : enhancement
1728 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1731 Frequency : rare, need acl's on inode.
1733 Description: client can't handle ost additional correctly
1734 Details : if ost was added after client connected to mds client can have
1735 hit lnet_try_match_md ... to big messages to wide striped files.
1736 in this case need teach client to handle config events about add
1737 lov target and update client max ea size at that event.
1739 Severity : enhancement
1741 Description: Update to sles9 kernel-2.6.5-7.314.
1743 Severity : enhancement
1745 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1748 Frequency : Create a symlink file with a very long name
1750 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1751 Details : If there is no extra space in the request for early cancels,
1752 ldlm_req_handles_avail() returns 0 instead of a negative value.
1757 Description: mds is deadlocked
1758 Details : in rare cases, inode in catalog can have i_no less than have parent
1759 i_no, this produce wrong order for locking during open, and
1760 parallel unlink can be lock open. this need teach mds_open to grab
1761 locks in resource id order, not at parent -> child order.
1763 Severity : enhancement
1765 Description: Add /proc entry for import status
1766 Details : The mdc, osc, and mgc import directories now have
1767 an import directory that contains useful import data for debugging
1768 connection problems.
1770 Severity : enhancement
1772 Description: Re-disable certain /proc logging
1773 Details : Enable and disable client's offset_stats, extents_stats and
1774 extents_stats_per_process stats logging on the fly.
1777 Frequency : Only on FC kernels 2.6.22+
1779 Description: oops in statahead
1780 Details : Do not drop reference count for the dentry from VFS when lookup,
1781 VFS will do that by itself.
1783 Severity : enhancement
1785 Description: Generic /proc file permissions
1786 Details : Set /Proc file permissions in a more generic way to enable non-
1787 root users operate on some /proc files.
1791 Description: Hitting mdc_commit_close() ASSERTION
1792 Details : Properly handle request reference release in
1793 ll_release_openhandle().
1797 Frequency : only patchless client
1798 Description: add workaround for race between add/remove dentry from hash
1800 Severity : enhancement
1802 Description: Allow OST glimpses to return PW locks
1806 Description: LBUG when llog conf file is full
1807 Details : When llog bitmap is full, ENOSPC should be returned for plain
1812 Description: Prevent import from entering FULL state when server in recovery
1816 Description: service mount cannot take device name with ":"
1817 Details : Only when device name contains ":/" will mount treat it as
1823 Description: replace ptlrpcd with the statahead thread to interpret the async
1824 statahead RPC callback
1828 Frequency : on recovery
1829 Description: I/O failures after umount during fail back
1830 Details : if client reconnected to restarted server we need join to recovery
1831 instead of find server handler is changed and process self eviction
1832 with cancel all locks.
1834 Severity : enhancement
1836 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1840 Description: Kernel BUG tries to release flock
1841 Details : Lustre does not destroy flock lock before last reference goes
1842 away. So always drop flock locks when client is evicted and
1843 perform unlock regardless of successfulness of speaking to MDS.
1845 Severity : enhancement
1847 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1849 Severity : enhancement
1851 Description: Upcall on Lustre log has been dumped
1852 Details : Allow for a user mode script to be called once a Lustre log has
1853 been dumped. It passes the filename of the dumped log to the
1854 script, the location of the script can be specified via
1855 /proc/sys/lnet/debug_log_upcall.
1860 Description: avoid messages about idr_remove called for id that is not allocated
1861 Details : Move assigment s_dev for clustered nfs to end of initialization,
1862 for avoid problem with error handling.
1867 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1868 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1869 CONN_USED_HASH and this prodice warning when put connection again
1875 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1876 Details : release reference to stats when client disconnected, not
1877 when export destroyed for avoid races when client destroyed
1878 after main ost export.
1882 Description: more cleanup in mds_lov
1883 Details : add workaround for get valid ost count for avoid warnings about
1884 drop too big messages, not init llog cat under semphore which
1885 can be blocked on reconnect and break normal replay, fix access
1888 Severity : enhancement
1890 Description: Export bytes_read/bytes_write count on OSC/OST.
1894 Description: Early reply size mismatch, MGC loses connection
1895 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1896 the connect flags are properly negotiated.
1900 Description: Properly propagate oinfo flags from lov to osc for statfs
1901 Details : restore missing copy oi_flags to lov requests.
1905 Description: exports in /proc are broken
1906 Details : recreate /proc entries for clients when they reconnect.
1908 Severity : enhancement
1910 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1911 : l_getgroups(8), lst(8), routerstat(8)
1912 Details : included man pages for llobdstat(8), llstat(8),
1913 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1915 Severity : enhancement
1917 Description: Implement lustre ll_show_options method.
1919 Severity : enhancement
1921 Description: Update to SLES9 kernel-2.6.5-7.312.
1923 Severity : enhancement
1925 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1929 Description: exports in /proc are broken
1930 Details : recreate /proc entries for clients when they reconnect.
1934 Description: don't fail open with -ERANGE
1935 Details : if client connected until mds will be know about real ost count
1936 get LOV EA can be fail because mds not allocate enougth buffer
1941 Description: Resolve device initialization race
1942 Details : Prevent proc handler from accessing devices added to the
1943 obd_devs array but yet be intialized.
1945 Severity : enhancement
1947 Description: configure's --enable-quota should check the
1948 : kernel .config for CONFIG_QUOTA
1949 Details : configure is terminated if --enable-quota is passed but
1950 : no quota support is in kernel
1952 Severity : enhancement
1954 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1956 Severity : enhancement
1958 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1962 Frequency : rare, on PPC clients
1963 Description: don't swab ost objects in response about directory, because
1965 Details : bug similar bug 14856, but in different function.
1967 Severity : enhancement
1969 Description: lfs quota tool enhancement
1970 Details : added units specifiers support for setquota, default to
1971 current uid/gid for quota report, short quota stats by
1972 default, nonpositional parameters for setquota, added
1973 llapi_quotactl manual page.
1975 Severity : enhancement
1977 Description: *optional* service tags registration
1978 Details : if the "service tags" package is installed on a Lustre node
1979 When the filesystem is mounted, a local-node service tag will
1980 be created. See http://inventory.sun.com/ for more information
1981 about the Service Tags asset management system.
1985 Description: Client runs out of low memory
1986 Details : Consider only lowmem when counting initial number of llap pages
1989 Frequency : occasional
1991 Description: add refcount for osc callbacks, so avoid panic on shutdown
1993 Severity : enhancement
1995 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1998 Frequency : testing only
2000 Description: sanity test 65a fails if stripecount of -1 is set
2001 Details : handle -1 striping on filesystem in ll_dirstripe_verify
2004 Frequency : only in unusual configurations
2006 Description: Kernel panic with find ost index.
2007 Details : lov_obd have panic if some OST's have sparse indexes.
2009 Severity : enhancement
2011 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
2014 Frequency : rarely, if filesystem is mounted with -o flock
2016 Description: do not process already freed flock
2017 Details : flock can possibly be freed by another thread before it reaches
2018 to ldlm_flock_completion_ast.
2021 Frequency : rarely, if filesystem is mounted with -o flock
2023 Description: LBUG during stress test
2024 Details : Need properly lock accesses the flock deadlock detection list.
2027 Frequency : rarely, if binaries are being run from Lustre
2029 Description: oops in page fault handler
2030 Details : kernel page fault handler can return two special 'pages' in
2031 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
2034 Frequency : rarely, during shutdown
2036 Description: timeout with invalidate import.
2037 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
2038 should be handled by ptlrpcd. This produce long age waiting and
2039 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
2045 Description: ASSERTION(CheckWriteback(page,cmd)) failed
2046 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
2050 Frequency : only with broken builds/installations
2052 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
2053 Details : just return an error to a user, put a console error message
2055 Severity : enhancement
2057 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
2059 Severity : enhancement
2061 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
2063 Severity : enhancement
2065 Description: enable MGS and MDT services start separately
2066 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
2067 a co-located MGS without starting the MGS, which is a complement
2068 to 'nosvc' mount option.
2071 Frequency : always, on big-endian systems
2073 Description: cleanup in ptlrpc code, related to PPC platform
2074 Details : store magic in native order avoid panic's in recovery on PPC
2075 node and forbid from this error in future. Also fix posibily
2076 of twice swab data. Fix get lov striping to userpace.
2079 Frequency : rarely, if replay get lost on server
2081 Description: server incorrectly drop resent replays lead to recovery failure.
2082 Details : do not drop replay according to msg flags, instead we check the
2083 per-export recovery request queue for duplication of transno.
2086 Frequency : after recovery
2088 Description: precreate to many object's after del orphan.
2089 Details : del orphan st in oscc last_id == next_id and this triger growing
2090 count of precreated objects. Set flag LOW to skip increase count
2091 of precreated objects.
2094 Frequency : rare, on clear nid stats
2096 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
2097 Details : when clean nid stats sometimes try destroy live entry,
2098 and this produce panic in free.
2101 Frequency : occasionally since 1.6.4
2103 Description: Stack overflow during MDS log replay
2104 Details : ease stack pressure by using a thread dealing llog_process.
2107 Frequency : very rare
2109 Description: MDT cannot be unmounted, reporting "Mount still busy"
2110 Details : Mountpoint references were being leaked during open reply
2111 reconstruction after an MDS restart. Drop mountpoint reference
2112 in reconstruct_open() and free dentry reference also.
2117 Description: wait until IO finished before start new when do lock cancel.
2118 Details : VM protocol want old IO finished before start new, in this case
2119 need wait until PG_writeback is cleared until check dirty flag
2120 and call writepages in lock cancel callback.
2125 Description: mds_mfd_close() ASSERTION(rc == 0)
2126 Details : In mds_mfd_close(), we need protect inode's writecount change
2127 within its orphan write semaphore to prevent possible races.
2130 Frequency : rare, on shutdown ost
2132 Description: don't hit live lock with umount ost.
2133 Details : shrink_dcache_parent can be in long loop with destroy dentries,
2134 use shrink_dcache_sb instead.
2137 Frequency : only when echo_client is used
2139 Description: don't panic with use echo_client
2140 Details : echo client pass NULL as client nid pointer and this produce NULL
2141 pointer dereference.
2144 Frequency : Always on 32-bit PowerPC systems
2146 Description: fix build on PPC32
2147 Details : compile code with -m64 flag produce wrong object file for PPC32.
2152 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2153 Details : In reconstruct_* functions, LASSERTs on both the data supplied
2154 by a client, and the data on disk are dangerous and incorrect.
2155 Change them with client eviction.
2157 Severity : enhancement
2159 Description: skiplist implementation simplification
2160 Details : skiplists are used to group compatible locks on granted list
2161 that was implemented as tracking first and last lock of each lock
2162 group the patch changes that to using doubly linked lists
2166 Description: delete compatibility for 32bit qdata
2167 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
2168 qunit. That means servers of b1_4 and servers of b1_8 can't be
2169 used together if users want to use quota.
2172 Frequency : only with administrator action
2174 Description: mount failure if config log has invalid conf_param setting
2175 Details : If administrator specified an incorrect configuration parameter
2176 with "lctl conf_param" this would cause an error during future
2177 client mounts. Instead, ignore the bad configuration parameter.
2180 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
2182 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
2183 Details : If blocks per group is less than blocksize*8, set rest of the
2187 Frequency : Application do stride read on lustre
2189 Description: The read performance will drop a lot if the application does
2191 Details : Because the stride_start_offset are missing in stride read-ahead,
2192 it will cause clients read a lot of unused pages in read-ahead,
2193 then the read-performance drops.
2197 Description: more ldlm soft lockups
2198 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2199 starve other threads from the resource lock for a long time in
2200 case of long waiting queue, so change the debug level from
2201 D_OTHER to the less frequently used D_INFO.
2203 Severity : enhancement
2205 Description: add -gid, -group, -uid, -user options to lfs find
2207 Severity : enhancement
2209 Description: ll_recover_lost_found_objs - recover objects in lost+found
2210 Details : OST corruption and subsequent e2fsck can leave objects in the
2211 lost+found directory. Using the "ll_recover_lost_found_objs"
2212 tool, these objects can be retrieved and data can be salvaged
2213 by using the object ID saved in the fid EA on each object.
2218 Description: this bug _only_ happens when inode quota limitation is very low
2219 (less than 12), so that inode quota unit is 1 at initialization.
2220 Details : if remaining quota equates 1, it is a sign to demonstate that quota
2221 is effective now. So least quota qunit should be 2.
2225 Description: Hung threads in invalidate_inode_pages2_range
2226 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
2227 one is completed. As a result, some RPCs are stuck in the queue
2232 Description: Procfs and llog threads access destoryed import sometimes.
2233 Details : Sync the import destoryed process with procfs and llog threads by
2234 the import refcount and semaphore.
2238 Description: mds fails to respond, threads stuck in ldlm_completion_ast
2239 Details : Sort source/child resource pair after updating child resource.
2244 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2245 Details : If insertion of an extent fails, then discard the inode
2246 preallocation and free data blocks else it can lead to duplicate
2251 Description: don't always update ctime in ext3_xattr_set_handle()
2252 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
2253 In some cases the ctime should not be updated, for example for
2254 2.0->1.8 compatibility it is necessary to delete an xattr and it
2255 should not update the ctime.
2259 Description: add quota statistics
2260 Details : 1. sort out quota proc entries and proc code.
2261 2. add quota statistics
2266 Description: quotas are not honored with O_DIRECT
2267 Details : all writes with the flag O_DIRECT will use grants which leads to
2268 this problem. Now using OBD_BRW_SYNC to guard this.
2272 Bugzilla : 15713/16362
2273 Description: Assertion in iopen_connect_dentry in 1.6.3
2274 Details : looking up an inode via iopen with the wrong generation number can
2275 populate the dcache with a disconneced dentry while the inode
2276 number is in the process of being reallocated. This causes an
2277 assertion failure in iopen since the inode's dentry list contains
2278 both a connected and disconnected dentry.
2282 Description: assertion failure in ldlm_handle2lock()
2283 Details : fix a race between class_handle_unhash() and class_handle2object()
2284 introduced in lustre 1.6.5 by bug 13622.
2286 Severity : enhancement
2288 Description: superblock lock contention with many SMP cores on one client
2289 Details : several client filesystem locks were highly contended on SMP
2290 NUMA systems with 8 or more cores. Per-CPU datastructures
2291 and more efficient locking implemented to reduce contention.
2296 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2297 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2302 Description: oops in sd_iostats_seq_show()
2303 Details : unloading/reloading the scsi low level driver triggers a kernel
2304 bug when trying to access the sd iostat file.
2309 Description: Kernel panics during QLogic driver reload
2310 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2311 patch, causing memory corruption.
2316 Description: journal_dev option does not work in b1_6
2317 Details : pass mount option during pre-mount.
2319 Severity : enhancement
2321 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2322 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2323 extent information of a file. It can be used to map logical blocks
2324 in a file to physical blocks in the block device.
2327 Frequency : only with adaptive timeout enabled
2329 Description: DEBUG_REQ() bad paging request
2330 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2331 because it can be accessed by reply_in_callback() without the
2335 Frequency : only on Cray X2
2337 Description: X2 build failures
2338 Details : fix build failures on Cray X2.
2342 Description: xid & resent requests
2343 Details : Initialize RPC XID from clock at startup (randomly if clock is
2348 Description: quota recovery deadlock during mds failover
2349 Details : This patch includes att18982, att18236, att18237 in bz14840.
2351 1. fix osts hang when mds does failover with quotaon
2352 2. prevent watchdog storm when osts threads wait for the
2357 Description: kernel panic on racer
2358 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2360 Severity : enhancement
2362 Description: Add lustre_start utility to start or stop multiple Lustre servers
2367 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2368 Details : In case of memory pressure, list_del() can be called twice on
2369 req->rq_history_list, causing a kernel oops.
2373 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2374 Details : fix stack overflow in the distributed lock manager by defering
2375 export eviction after a failed ast to the elt thread instead of
2376 handling it in the dlm interpret routine.
2378 Severity : enhancement
2380 Description: More exported tunables for mballoc
2381 Details : Add support for tunable preallocation window and new tunables for
2382 large/small requests
2386 Description: Detect corruption of block bitmap and checking for preallocations
2387 Details : Checks validity of on-disk block bitmap. Also it does better
2388 checking of number of applied preallocations. When corruption is
2389 found, it turns filesystem readonly to prevent further corruptions.
2393 Frequency : only for big-endian servers
2394 Description: Check for big-endian system while mounting fs with extents feature
2395 Details : Mounting a filesystem with extents feature will fail on big-endian
2396 systems since ext3-based ldiskfs is not supported on big-endian
2397 systems. Can be over-ridden with "bigendian_extents" mount option.
2401 Description: Excessive recovery window
2402 Details : With AT enabled, the recovery window can be excessively long (6000+
2403 seconds). To address this problem, we no longer use
2404 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2405 timeout no longer depends on the service time, it is set to
2406 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2407 time via pb_service_time.
2410 Descriptoin: Don't sync journal after every i/o
2411 Details : Implement write RPC replay to allow server replies for write RPCs
2412 before data is on disk. However, this feature is disabled by
2413 default since some issues leading to data corruptions have been
2414 found during recovery (e.g. bug 19128). This feature can be enabled
2415 by running the following command on the OSSs:
2416 lctl set_param obdfilter.*.sync_journal=0
2420 Description: Watchdog triggered on MDS failover
2421 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2422 the OSTs know that the MDS "UUID" can be reused for the same export
2423 from a different NID, so we do not need to wait for the export to
2428 Descriptoin: Lustre detected file system corruption with inode out of bounds
2429 Details : don't update i_size on MDS_CLOSE for directories. This causes
2430 directory corruptions on the MDT.
2432 -------------------------------------------------------------------------------
2434 2008-05-26 Sun Microsystems, Inc.
2436 * Support for kernels:
2437 2.6.5-7.311 (SLES 9),
2438 2.6.9-67.0.7.EL (RHEL 4),
2439 2.6.16.54-0.2.5 (SLES 10),
2440 2.6.18-53.1.14.el5 (RHEL 5),
2441 2.6.22.14 vanilla (kernel.org)
2442 * Client support for unpatched kernels:
2443 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2444 2.6.16 - 2.6.22 vanilla (kernel.org)
2445 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2446 we do not recommend using patchless RHEL4 clients with kernels
2447 prior to 2.6.9-55EL (RHEL4U5).
2448 * Recommended e2fsprogs version: 1.40.7-sun1
2449 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2450 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2451 removed cwd "./" (refer to Bugzilla 14399).
2452 * A new quota file format has been introduced in 1.6.5.
2453 The format conversion from prior releases is handled transparently,
2454 but releases older than 1.4.12/1.6.5 will not understand this new
2455 format. The automatic format conversion can be avoided by running
2456 the following command on the MDS before upgrading:
2457 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2458 For more information, please refer to bugzilla 13904.
2462 Description: quota performance fix
2463 Details : quota data is written in journalled mode instead of ordered to
2464 increase performance
2468 Description: lfs support for human-readable quota grace time strings
2469 Details : lfs setquota -t and lfs quota -t represent quota grace times
2470 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2473 Frequency : always with o2ib 1.3 and sles10
2475 Description: fix build with SLES10 and o2ib v3.
2476 Details : sles10 uses diffrent name for Module.symver file but configure
2477 assume this file has same name on RHEL/SLES/vanila kernels.
2480 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2482 Description: MDS may lose file striping (and hence file data) in some cases
2483 Details : If there are additional extended attributes stored on the MDS,
2484 in particular ACLs, SELinux, or user attributes (if user_xattr
2485 is specified for the client mount options) then there is a risk
2486 of attribute loss. Additionally, the Lustre file striping
2487 needs to be larger than default (e.g. striped over all OSTs),
2488 and an additional attribute must be stored initially in the
2489 inode and then increase in size enough to be moved to the
2490 external attribute block (e.g. ACL growing in size) for file
2493 Severity : enhancement
2495 Description: add message levels for liblustreapi
2500 Description: MDT cannot be unmounted, reporting "Mount still busy"
2501 Details : Mountpoint references were being leaked during open reply
2502 reconstruction after an MDS restart. Drop mountpoint reference
2503 in reconstruct_open() and free dentry reference also.
2508 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2509 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2510 will result in an -EAGAIN and in the exit call path we call
2511 alloc_rr() which will with increasing aggressiveness attempt to
2512 aquire precreated objects on the minimum number of required OSCs.
2516 Description: Use old size assignment to avoid deadlock
2517 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2518 the scheduling while holding a spinlock. We do not need locking
2519 for size in ll_update_inode() because size is only updated from
2520 the MDS for directories or files without objects, so there is no
2521 other place to do the update, and concurrent access to such inodes
2522 are protected by the inode lock.
2526 Description: Use __u64 instead of int for valid bits
2530 Description: resolve "_IOWR redefined" build error on SLES10
2534 Description: dump the memory debugging after all modules are unloaded to
2535 suppress false negative in conf_sanity test 39
2539 Description: the recovery timer never expires
2540 Details : for new client connect request, the recovery timer should not be
2541 reset, otherwise recovery timer will never expired, if the old
2542 client never come. Only old client connect and first connection
2543 req should trigger recovery timer reset.
2547 Description: the min numbers of lproc stats are wrong
2548 Details : adding a new constant LC_MIN_INIT and use it for initialization
2552 Frequency : always with interactive lfs
2554 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2557 Frequency : with multiple concurrent readdir processes in same directory
2558 Bugzilla : 15406, 15169, 15175
2559 Description: misc fixes for directory readahead.
2560 Details : prevent previous statahead async RPC callback from processing the
2561 current "statahead_info", race condition between async RPC callback
2562 add dentry into dentry hash table and "ls" thread revalidate such
2563 dentry, statahead his/miss control for hidden items, and so on.
2565 Severity : enhancement
2567 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2571 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2572 Details : several u32 variables declarations are replaced with u64 declarations
2577 Description: fix SLES kernel versioning
2578 Details : the kernel version for our SLES 10 kernel did not include a "-"
2579 before the "smp" at the end. while this was not a problem in
2580 general, it did mean that software trying to use the kernel
2581 version to try to detect a vendor specific kernel would fail.
2582 this was most evident by the OFED build scripts.
2587 Description: Don't update lov_desc members until making sure they are valid
2588 Details : When updating lov_desc members via proc fs, need fix their
2589 validities before doing the real update.
2592 Frequency : very rare
2594 Description: don't put request into delay list while invalidate in flight.
2595 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2596 import in flight. this produce timeout for invalidate and sometimes
2597 can cause stale data.
2599 Severity : enhancement
2601 Description: Update kernel to SLES9 2.6.5-7.311.
2603 Severity : enhancement
2605 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2610 Frequency : on PPC only
2611 Description: not convert ost objects for directory because it's not exist.
2612 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2614 Severity : enhancement
2616 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2617 __u64 as usingied long long type.
2620 Frequency : rare, on shutdown
2622 Description: race process ast vs remove callback
2623 Details : removing callback before disconnect import open race with
2624 processing callback.
2626 Severity : enhancement
2628 Description: Update kernel to SLES9 2.6.5-7.311.
2630 Severity : enhancement
2632 Description: Files open for execute are not marked busy on SLES10
2633 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2635 Severity : enhancement
2637 Description: Add server support for vanilla-2.6.22.14.
2640 Frequency : occasional
2642 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2643 Details : If an OST is down the MDS will hang indefinitely in
2644 obd_statfs_rqset() waiting for the statfs data. While for
2645 MDS QOS usage of statfs, it should not stuck in waiting.
2647 Severity : enhancement
2649 Description: Disable adaptive timeouts by default
2652 Frequency : on network error
2654 Description: panic with double free request if network error
2655 Details : mdc_finish_enqueue is finish request if any network error occuring,
2656 but it's true only for synchronus enqueue, for async enqueue
2657 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2661 Frequency : rare, on recovery
2663 Description: read procfs can produce deadlock in some situation
2664 Details : Holding lprocfs lock with send rpc can produce block for destroy
2665 obd objects and this also block reconnect with -EALREADY.
2666 This isn't fix all lprocfs bugs - but make it rare.
2668 Severity : enhancement
2670 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2673 Frequency : frequent on X2 node
2675 Description: mdc_set_open_replay_data LBUG
2676 Details : Set replay data for requests that are eligible for replay.
2681 Description: lustre_mgs: operation 101 on unconnected MGS
2682 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2683 MGC, and late on MGC cannot successfully connect to MGS and a lot
2684 of the error messages complaining that MGS is not connected.
2687 Frequency : rare, depends on device drivers and load
2689 Description: MDS or OSS nodes crash due to stack overflow
2690 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2691 In some cases, in conjunction with device drivers that use a lot
2692 of stack, the MDS (or possibly OSS) service threads could overflow
2693 the stack. One change which was identified to consume additional
2694 stack has been reworked to avoid the extra stack usage.
2696 Severity : enhancement
2698 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2700 Severity : enhancement
2702 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2704 Severity : enhancement
2706 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2708 Serverity : enhancement
2710 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2713 Frequency : rare on shutdown OST
2715 Description: Don't allow skipping OSTs if index has been specified.
2716 Details : Don't allow skipping OSTs if index has been specified, make locking
2717 in internal create lots better.
2722 Description: ASSERTION(!PageDirty(page)) failed
2723 Details : Wrong check could lead to an assertion failure under specific
2729 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2730 Details : access to bitfield in structure is always rounded to long
2731 and this produce problem with not atomic change any bit.
2736 Description: Lustre make rpms failed.
2737 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2738 builds Lustre rpms from tarball.
2740 Severity : enhancement
2742 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2745 Frequency : rare on shutdown OST
2747 Description: If llog cancel was not send before clean_exports phase, this can
2748 produce deadlock in llog code.
2749 Details : If llog thread has last reference to obd and call class_import_put
2750 this produce deadlock because llog_cleanup_commit_master wait when
2751 last llog_commit_thread exited, but this never success because was
2752 called from llog_commit_thread.
2755 Frequency : only if OST index is skipped
2757 Description: NULL lov_tgts causing MDS oops
2758 Details : more safe checks for NULL lov_tgts for avoid oops.
2760 Severity : enhancement
2762 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2764 Severity : enhancement
2766 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2771 Description: make mgs_setparam() handle fsname containing dash
2772 Details : fsname containing a dash does not work with lctl conf_param
2774 Severity : enhancement
2776 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2779 Frequency : rare, in recovery and (or) destroy lovobjid file.
2781 Description: rewrite lov objid code.
2782 Details : Cleanup for lov objid code, remove scability problems and wrong
2783 locking. Fix sending last_id into ost.
2785 Severity : enhancement
2787 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2789 Severity : enhancement
2791 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2792 Details : Use d_move_locked instead of __d_move.
2795 Frequency : rare, at shutdown
2797 Description: access already free / zero obd_namespace.
2798 Details : if client_disconnect_export was called without force flag set,
2799 and exist connect request in flight, this can produce access to
2800 NULL pointer (or already free pointer) when connect_interpret
2801 store ocd flags in obd_namespace.
2804 Frequency : only at startup
2806 Description: not alloc memory with spinlock held.
2807 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2808 if any spinlock held.
2813 Description: lfs find does not continue on file error
2814 Details : Continue other files processing when a file/dir is absent.
2818 Description: Inconsistent usage of lustre_pack_reply()
2819 Details : Standardize the usage of lustre_pack_reply() such that it
2820 always generate a CERROR on failure.
2823 Frequency : very rare
2825 Description: Fix replay if there is an un-replied request and open
2826 Details : In some cases, older replay request will revert the
2827 mcd->mcd_last_xid on MDS which is used to record the client's
2828 latest sent request.
2830 Severity : enhancement
2832 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2834 Severity : enhancement
2836 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2838 Severity : enhancement
2840 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2842 Severity : enhancement
2844 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2846 Severity : enhancement
2848 Description: Build SLES10 patchless client fails
2849 Details : The configure was broken by run ./configure with
2850 --with-linux-obj=.... argument for patchless client. When the
2851 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2852 can't search header adequately. Use absolute path such as
2853 -I($LINUX)/include instead.
2855 Severity : enhancement
2857 Description: Lustre Page Accounting
2858 Details : New macros for page alloc and free which enable accounting
2859 of page allocation of Lustre. Use percpu counters to store memory
2860 and page statistics.
2863 Frequency : only if debugging is disabled
2865 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2866 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2867 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2868 these macros and replace them with their definitions instead.
2873 Description: interrupt oig_wait produce painc on resend.
2874 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2875 private set, and this produce situation when rq_ptlrpcd_data not
2876 copyed to new allocated request and triggered LBUG on assert
2877 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2878 ptlrpcd set if request is changed and can be safety changed to use
2881 Severity : enhancement
2883 Description: organize the server-side client stats on per-nid basis
2884 Details : Change the structure of stats under obdfilter and mds to
2892 The "uuid"s file would list the uuids of _active_ exports.
2893 And the clear entry is to clear all stats and stale nids.
2898 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2899 a full node pseudo-hang
2900 Details : Concurrent access to the same directory from multiple clients with
2901 intensive file creation/removal can cause a client node to spin in
2902 ll_readdir(). i_version must be increased every time the lock is
2903 cancelled to ensure a revalidate is done.
2908 Description: touch file failed when fs is not full
2909 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2910 otherwise we can get ENOSP while fs is not full.
2913 Frequency : only for Cray XT3
2914 Bugzilla : 12829/13455
2915 Description: Changing primary group doesn't change the group lustre assigns to
2917 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2918 overridden with the primary group provided by the group upcall,
2919 whereas the supplied fsgid can be trusted if it is in the list of
2920 supplementary groups returned by the group upcall.
2922 Severity : enhancement
2924 Description: Root Squash Functionality
2925 Details : Implementation of NFS-like root squash capability. Specifically,
2926 don't allow someone with root access on a client node to be able
2927 to manipulate files owned by root on a server node.
2929 Severity : enhancement
2931 Description: Slow trucate/writes to huge files at high offsets.
2932 Details : Directly associate cached pages to lock that protect those pages,
2933 this allows us to quickly find what pages to write and remove
2934 once lock callback is received.
2939 Description: Too many locks accumulating on client during NFS usage
2940 Details : mds_open improperly used accmode to find out access mode to a
2941 file. Also mdc_intent_lock logic to find out if we already have
2942 lock similar to just received was flawed since introduction of
2943 skiplists - locks are now added to the front of the granted
2948 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2950 Details : In obd setup failure handler, obd_stopping will not necessarily to
2951 be 1, and obd_set_up should also be checked to make sure whether
2952 obd is completely setup.
2954 Severity : enhancement
2956 Description: Allow masking D_WARNING, D_ERROR messages from console
2957 Details : Console messages can now be disabled via lnet.printk.
2962 Description: User code with malformed file open parameter crashes client node
2963 Details : Before packing join_file req, all the related reference should be
2964 checked carefully in case some malformed flags cause fake join_file
2970 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2971 grain of quota doing harm to quota's accuracy
2972 Details : qunit size will be changed when quota limitation is too low/high;
2973 record the pending quota write in order to get more accureate
2974 quota; delete the patch for bug12588, which is unnecessary when
2975 this patch is landed. This bug also contains fixes for bug 14526,
2976 14299, 14601 and 13794.
2980 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2981 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2982 mode is equal to the granted mode. However, it should also check
2983 that LDLM_FL_CP_REQD is not set.
2987 Description: Heavy nfs access might result in deadlocks
2988 Details : After ELC code landed, it is now improper to enqueue any mds
2989 locks under och_sem, because enqueue might want to decide to
2990 cancel open locks for same inode we are holding och_sem for.
2994 Description: 35% write performance drop with ldiskfs2 when quotas are on
2995 Details : Enable ext3 journalled quota by default to improve performance
2996 when quotas are turned on.
3000 Description: Client eviction while running blogbench
3001 Details : A lot of unlink operations with concurrent I/O can lead to a
3002 deadlock causing evictions. To address the problem, the number of
3003 oustanding OST_DESTROY requests is now throttled to
3004 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
3005 callbacks are processed in priority.
3008 Frequency : RHEL4 only
3010 Description: mkfs is very slow on IA64/RHEL4
3011 Details : A performance regression has been discovered in the MPT Fusion
3012 driver between versions 3.02.73rh and 3.02.99.00rh. As a
3013 consequence, we have downgraded the MPT Fusion driver in the RHEL4
3014 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
3017 Frequency : PPC/PPC64 only
3019 Description: conflicts between asm-ppc64/types.h and lustre_types.h
3020 Details : fix duplicated definitions between asm-ppc64/types.h and
3021 lustre_types.h on PPC.
3024 Frequency : PPC/PPC64 only
3026 Description: asm-ppc/segment.h does not exist
3027 Details : fix compile issue on PPC.
3031 Description: data checksumming impacts single node performance
3032 Details : add support for several checksum algorithms. Currently, CRC32 and
3033 Adler-32 are supported. The checksum type can be changed on the fly
3034 through /proc/fs/lustre/osc/*/checksum_type.
3038 Description: use adler32 for page checksums
3039 Details : when available, use the Adler-32 algorithm instead of CRC32 for
3044 Description: better handle error messages in extents code
3046 Severity : enhancement
3048 Description: SNMP support enhancement
3049 Details : Adding total number of sampled request for an MDS node in snmp
3052 Severity : enhancement
3054 Description: Optimize ldlm waiting list processing for PR extent locks
3055 Details : When processing waiting list for read extent lock and meeting read
3056 lock that is same or wider to it that is not contended, skip
3057 processing rest of the list and immediatelly return current
3058 status of conflictness, since we are guaranteed there are no
3059 conflicting locks in the rest of the list.
3063 Description: Time out and refuse to reconnect
3064 Details : When the failover node is the primary node, it is possible
3065 to have two identical connections in imp_conn_list. We must
3066 compare not conn's pointers but NIDs, otherwise we
3067 can defeat connection throttling.
3071 Description: Client not clear own cache if answer to reconnect is lost.
3072 Details : Client gets evicted from server. Now client also thinks it is
3073 disconnected (or gets ENOTCONN on its operation) and decides to
3074 reconnect. Server receives reconnect message, but cannot find
3075 export. New export is created that is fully valid (new cookie!),
3076 but reply is lost and not reported to client. Client reconnects
3077 again and gets back a just-created connection, but it is not new
3078 so client thinks it was not evicted and does not do recovery.
3082 Description: Detect stride IO mode in read-ahead
3083 Details : When a client does stride read, read-ahead should detect that and
3084 read-ahead pages according to the detected stride pattern.
3088 Description: build for x2 fails
3089 Details : fix compile issue on Cray systems.
3091 Severity : enhancement
3093 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
3094 Details : This greatly improves speed of NFS writes on 2.6 kernels.
3097 Frequency : only on PPC/SLES10
3099 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
3100 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
3101 be defined. Add a hack in mkfs_lustre.c to work around this compile
3106 Description: LASSERT on MDS when client holding flock lock dies
3107 Details : ldlm pool logic depends on number of granted locks equal to
3108 number of released locks which is not true for flock locks, so
3109 just exclude such locks from consideration.
3113 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
3114 Details : Use fsfilt_sync() for both the whole filesystem sync and
3115 individual file sync to eliminate dangerous inode locking
3116 with I_LOCK that can lead to a deadlock.
3120 Description: Update an obsolete wirecheck.c generator
3121 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
3125 Description: Client can panic on open sometimes
3126 Details : It is possible that we try to free already freed request in
3127 ll_file_open in some error cases when we send request from
3132 Description: performance in 1.6.3
3133 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
3134 sure that 1MB requests are not fragmented by the block layer.
3138 Description: LDLM soft lockups - improvement
3139 Details : It is be possible to send the lock handle along with each read
3140 or write request because the client is already doing a lock match
3141 itself so there isn't any reason the OST should have to re-do that
3147 Description: lfs quota fails with deactivated OSTS
3148 Details : With this patch, three improvements are included:
3149 1. detete the softlimit in mds and osts when use "lfs quota".
3150 2. display the inaccurate data in the output of "lfs quota".
3151 3. try to get quota info when "lfs quota" is executed.
3156 Description: Extent locks not granted with no conflicts sometimes.
3157 Details : When race occurs in glimpse handler and nothing is returned,
3158 we do not reprocess the queue after lock cancel, and that leads
3159 to a stall until next activity on a resource
3162 Frequency : failover with quotaon
3164 Description: during mds failovers with quota on, OSTs got into deadlock state
3165 and causing dumpstack.
3166 Details : for every quota slave, at any time, there is only one quota req
3167 is sent to quota master for every uid/gid. Before that quota req
3168 returns, all the thread relative to the same uid/gid will wait.
3169 So if the quota req is lost because mds failovers or any other
3170 reasons, this bug will be hit. Now, dqacq_interpret() will handle
3171 quota reqs who time out.
3173 Severity : enhancement
3176 Description: when quota slave checks if quota is enough, there is an unnecessary
3178 Details : place this wait on necessary place instead of always waiting.
3180 --------------------------------------------------------------------------------
3182 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
3184 * Support for kernels:
3185 2.6.5-7.286 (SLES 9),
3186 2.6.9-55.0.9.EL (RHEL 4),
3187 2.6.16.53-0.8 (SLES 10),
3188 2.6.18-8.1.14.el5 (RHEL 5),
3189 2.6.18.8 vanilla (kernel.org)
3190 * Client support for unpatched kernels:
3191 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3192 2.6.16 - 2.6.22 vanilla (kernel.org)
3193 * Due to recently discovered recovery problems, we do not recommend
3194 using patchless RHEL 4 clients with this or any earlier release.
3195 * Recommended e2fsprogs version: 1.40.2-cfs1
3196 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3199 Frequency : occasional
3201 Description: MDS hang or stay in waiting lock
3202 Details : If client receive lock with CBPENDING flag ldlm need send lock
3203 cancel as separate rpc, to avoid situation when cancel request
3204 can't processed due all i/o threads stay in wait lock.
3207 Frequency : occasional
3209 Description: Do not fail import if osc_interpret_create gets -EAGAIN
3210 Details : If osc_interpret_create got -EAGAIN it immediately exits and
3211 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
3212 oscc_has_objects and see OSC has no objests and call
3213 oscc_internal_create to resend create request.
3215 Severity : enhancement
3217 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
3218 Details : Update which_patch & target file for SLES10 latest kernel.
3220 Severity : enhancement
3222 Description: add --type and --size parameters to lfs find
3223 Details : Enhance lfs find by adding filetype and filesize parameters. Also
3224 multiple OBDs can now be specified for the --obd option.
3226 Severity : enhancement
3228 Description: eliminate client locks in face of contention
3229 Details : file contention detection and lockless i/o implementation
3230 for contended files.
3232 Severity : enhancement
3234 Description: Remove client patches from SLES 10 kernel.
3235 Details : This causes SLES 10 clients to behave as patchless clients
3236 even on a Lustre-patched (server) kernel.
3238 Severity : enhancement
3240 Description: use i_size_read and i_size_write in 2.6 port
3241 Details : replace inode->i_size access with i_size_read/write()
3244 Frequency : when removing large files
3246 Description: scheduling issue during removal of large Lustre files
3247 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
3248 It causes scheduling issues when removing large files (17TB in the
3254 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
3255 Details : Impossible (EPROTO is returned) to access a directory that has a
3256 non-default striping and ACLs.
3259 Frequency : only on ppc
3261 Description: /proc/fs/lustre/devices broken on ppc
3262 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3263 We should make sure the type of 'index' is loff_t and then cast
3264 explicitly as needed below. Do not assign an explicitly cast
3268 Frequency : only for rhel5
3270 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3271 Details : Modify the target file & which_kernel.
3274 Frequency : if the uninit_groups feature is enabled on ldiskfs
3276 Description: e2fsck reports "invalid unused inodes count"
3277 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3278 feature and only a single inode is created in a group then the
3279 "bg_unused_inodes" count is incorrectly updated. Creating a
3280 second inode in that group would update it correctly.
3285 Description: buffer overruns could theoretically occur
3286 Details : llapi_semantic_traverse() modifies the "path" argument by
3287 appending values to the end of the origin string, and a buffer
3288 overrun may occur. Adding buffer overrun check in liblustreapi.
3290 Severity : enhancement
3292 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3295 Frequency : only if filesystem is inconsistent
3297 Description: handle "serious error: objid * already exists" more gracefully
3298 Details : If LAST_ID value on disk is smaller than the objects existing in
3299 the O/0/d* directories, it indicates disk corruption and causes an
3300 LBUG(). If the object is 0-length, then we should use the existing
3301 object. This will help to avoid a full fsck in most cases.
3303 Severity : enhancement
3305 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3306 Details : Modify vm-tunables-rhel4.patch.
3308 Severity : enhancement
3310 Description: Kernel config for 2.6.18-vanilla.
3311 Details : Modify targets/2.6-vanilla.target.in.
3312 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3313 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3314 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3315 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3318 Frequency : occasional
3320 Description: improve handling recoverable errors
3321 Details : If request processed with error which can be recoverable on server
3322 request should be resend, otherwise page released from cache and
3328 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3329 Details : Modify target file & which_patch.
3330 A flaw was found in the IA32 system call emulation provided
3331 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3332 value could be stored in the %RAX register, which could trigger an
3333 out-of-bounds system call table access. An untrusted local user
3334 could exploit this flaw to run code in the kernel
3335 (ie a root privilege escalation). (CVE-2007-4573).
3339 Description: change order of libsysio includes
3340 Details : '#include sysio.h' should always come before '#include xtio.h'
3342 Severity : enhancement
3344 Description: adapt the lustre_config script to support the upgrade case
3345 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3346 server targets to 1.6 in parallel.
3351 Description: To avoid grant space > avaible space when the disk is almost
3352 full. Without this patch you might see the error "grant XXXX >
3353 available" or some LBUG about grant, when the disk is almost
3355 Details : In filter_check_grant, for non_grant cache write, we should
3356 check the left space by if (*left > ungranted + bytes), instead
3357 of (*left > ungranted), because only we are sure the left space
3358 is enough for another "bytes", then the ungrant space should be
3359 increase. In client, we should update cl_avail_grant only there is
3360 OBD_MD_FLGRANT in the reply.
3365 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3366 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3367 described in CVE-2007-4573. This problem could allow untrusted
3368 local users to gain root access.
3371 Frequency : when using O_DIRECT and quotas
3373 Description: Incorrect file ownership on O_DIRECT output files
3374 Details : block usage reported by 'lfs quota' does not take into account
3375 files that have been written with O_DIRECT.
3379 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3380 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3381 window to 0. If it does hit cache miss, and read-pattern is still
3382 stride-io mode, does not reset the stride window, but also does not
3383 increase the stride window length in this case.
3385 --------------------------------------------------------------------------------
3387 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3389 * Support for kernels:
3390 2.6.5-7.286 (SLES 9),
3391 2.6.9-55.0.2.EL (RHEL 4),
3392 2.6.16.46-0.14 (SLES 10),
3393 2.6.18-8.1.8.el5 (RHEL 5),
3394 2.6.18.8 vanilla (kernel.org)
3395 * Client support for unpatched kernels:
3396 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3397 2.6.16 - 2.6.21 vanilla (kernel.org)
3398 * Due to recently discovered recovery problems, we do not recommend
3399 using patchless RHEL 4 clients with this or any earlier release.
3400 * Recommended e2fsprogs version: 1.40.2-cfs1
3401 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3405 Description: Fix errors in lfs documentation
3406 Details : Fixes man pages
3408 Severity : enhancement
3410 Description: Adaptive timeouts
3411 Details : RPC timeouts adapt to changing server load and network
3412 conditions to reduce resend attempts and improve recovery time.
3414 Severity : enhancement
3416 Description: llapi_file_create() does not allow some changes
3417 Details : add llapi_file_open() that allows specifying the file creation
3418 mode and open flags, and also returns an open file handle.
3420 Severity : enhancement
3422 Description: df doesn't work properly if diskfs blocksize != 4K
3423 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3426 Frequency : always on directories with default striping set
3428 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3429 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3431 Severity : enhancement
3433 Description: merge and cleanup kernel patches.
3434 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3436 Severity : enhancement
3438 Description: RedHat Update kernel for RHEL5
3439 Details : Kernel config file for RHEL5.
3441 Severity : enhancement
3443 Description: OSS needs mutliple precreate threads
3444 Details : Add ability to start more than one create thread per OSS.
3447 Frequency : only with quota on the root user
3449 Description: mds_obd_create error creating tmp object
3450 Details : When the user sets quota on root, llog will be affected and can't
3451 create files and write files.
3454 Frequency : Always on ia64 patchless client, and possibly others.
3456 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3457 Details : This allows the patchless client to be loaded on architectures
3458 without this export.
3463 Description: disorder of journal start and llog_add cause deadlock.
3464 Details : in llog_origin_connect, journal start should happen before
3465 llog_add keep the same order as other functions to avoid
3468 Severity : enhancement
3470 Description: RedHat Update kernel for RHEL5
3471 Details : Modify the kernel config file more closer RHEL5.
3473 Severity : enhancement
3475 Description: Build failure against Centos5 (RHEL5)
3476 Details : Define PAGE_SIZE when it isn't present.
3479 Frequency : occasionally when using NFS
3481 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3482 Details : with NFS, the anon dentry's parent was set to itself in
3483 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3484 corresponding dentry other than use rec->ur_name.
3486 Severity : enhancement
3488 Description: enable data checksumming by default
3489 Details : enable checksum by default, allow --disable-checksum
3490 configure option and "-o nochecksum" mount option. Checksums
3491 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3494 Frequency : Occasionally with failover
3496 Description: Client eviction due to failover config
3497 Details : after a connection loss, the lustre client should attempt to
3498 reconnect to the last active server first before trying the
3499 other potential connections.
3501 Severity : enhancement
3503 Description: client-side metadata stat-ahead during readdir(directory readahead)
3504 Details : perform client-side metadata stat-ahead when the client detects
3505 readdir and sequential stat of dir entries therein
3510 Description: when mds and osts use different quota unit(32bit and 64bit),
3511 quota will be released repeatly.
3512 Details : void sending multiple quota reqs to mds, which will keep the
3513 status between the reqs.
3516 Frequency : only with liblustre clients on XT3
3518 Description: evictions taking too long
3519 Details : allow llrd to evict clients directly on OSTs
3523 Description: osts not allocated evenly to files
3524 Details : change the condition to increase offset_idx
3526 Severity : enhancement
3528 Description: self-adjustable client's lru lists
3529 Details : use adaptive algorithm for managing client cached locks lru
3530 lists according to current server load, other client's work
3531 pattern, memory activities, etc. Both, server and client
3532 side namespaces provide number of proc tunables for controlling
3537 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3538 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3539 readdir code from ext3) from scratch preserving functionality.
3543 Description: Only those disconnect error should be returned by rq_status.
3544 Details : In open/enqueue processs, Some errors, which will cause client
3545 disconnected, should be returned by rq_status, while other
3546 errors should still be returned by intent, then mdc or llite will
3549 Severity : enhancement
3551 Description: Tune the kernel for good SCSI performance.
3552 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3553 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3557 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3558 Bugzilla : 13547, 13627
3559 Description: Data corruption for OSTs that are formatted larger than 2TB
3561 Details : When generating the bio request for lustre file writes the
3562 sector number would overflow a temporary variable before being
3563 used for the IO. The data reads correctly from Lustre (which
3564 will overflow in a similar manner) but other file data or
3565 filesystem metadata may be corrupted in some cases.
3569 Description: "lfs find -obd UUID" prints directories
3570 Details : "lfs find -obd UUID" will return all directory names instead
3571 of just file names. It is incorrect because the directories
3572 do not reside on the OSTs.
3575 Bugzilla : 13596, 13614
3576 Description: MDS hang after unclean shutdown of lots of clients
3577 Details : Never resend AST requests.
3581 Frequency : Always, for kernels after 2.6.16
3582 Description: Fix warning idr_remove called for id=.. which is not allocated.
3583 Details : Last kernels save old s_dev before kill super and not allow
3584 to restore from callback - restore it before call kill_anon_super.
3588 Description: Fix error on 'ls .' at the top of the Lustre mount.
3589 Details : Don't revalidate dentry if it is a root dentry.
3591 --------------------------------------------------------------------------------
3593 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3595 * Support for kernels:
3596 2.6.5-7.286 (SLES 9),
3597 2.6.9-55.0.2.EL (RHEL 4),
3598 2.6.16.46-0.14 (SLES 10),
3599 2.6.18-8.1.8.el5 (RHEL 5),
3600 2.6.18.8 vanilla (kernel.org)
3601 * Client support for unpatched kernels:
3602 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3603 2.6.16 - 2.6.22 vanilla (kernel.org)
3604 * Due to recently discovered recovery problems, we do not recommend
3605 using patchless RHEL 4 clients with this or any earlier release.
3606 * Recommended e2fsprogs version: 1.39.cfs8
3607 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3612 Description: block reactivating mgc import until all deactivates complete
3613 Details : Fix race when failing back MDT/MGS to itself (testing)
3615 Severity : enhancement
3617 Description: lfs setstripe enhancement
3618 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3621 Frequency : mds/oss recovery
3623 Description: llog ctxt is refrenced after it has been freed.
3624 Details : llog ctxt refcount was added to avoide the race between ctxt free
3625 and llog recovery process. Each llog user must hold ctxt refcount
3626 before it access the llog. And the llog ctxt can only be freed
3627 when its refcount is zero.
3629 Severity : enhancement
3631 Description: randomly memory allocation failure util
3632 Details : Make lustre randomly failed allocating memory for testing purpose.
3635 Frequency : only for SLES10
3637 Description: Update kernel patch for SLES10 SP1
3638 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3640 Severity : enhancement
3642 Description: omit set fsid for export NFS
3643 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3644 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3647 Frequency : after network failures
3649 Description: Add sync option to mount_lustre.c
3650 Details : Client loses data written to lustre after a network interruption.
3652 Severity : enhancement
3654 Description: Error message improvement.
3655 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3657 Severity : enhancement
3659 Description: don't use GFP_* in generic Lustre code.
3660 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3662 Severity : enhancement
3664 Description: obdclass is limited by single OBD_ALLOC(idarray)
3665 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3667 Severity : enhancement
3669 Description: updated patchess for new RHEL4 kernel
3670 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3672 Severity : enhancement
3674 Description: warnings with build patchless client with vanila 2.6.19 and up
3675 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3676 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3678 Severity : enhancement
3680 Description: O_DIRECT bypasses client statistics.
3681 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3682 read_ahead_stats were updated. Copied stats section from
3683 osc_send_oap_rpc() into async_internal().
3686 Frequency : only for Cray XT3
3688 Description: peer credits not enough on many OST per OSS systems.
3689 Details : Use new lnet way to add credits as we need those for pings and ASTs
3692 Frequency : only with liblustre
3694 Description: Liblustre is not releasing flock locks on file close.
3695 Details : Release flock locks on file close.
3698 Frequency : only for RHEL4
3700 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3701 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3702 Update target file and kernel config.
3706 Description: lustre support for RHEL5
3707 Details : Add support for RHEL5.
3712 Description: ASSERTION(export != NULL) failed in target_handle_connect
3713 Details : Assetion hit is result of rare race between disconnect and connet
3714 to same nid. target_handle_connect found old connect cockie and
3715 tried to reconnect, but can't find export for this cockie.
3720 Description: umount blocks forever on error
3721 Details : In result of wrong using obd_no_recov and obd_force flags client
3722 can hand if cancel or some other requests is lost.
3725 Frequency : Only for SLES
3727 Description: sanity_quota fail test_1
3728 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3729 file, which makes TSTID[2] inunique.
3731 Severity : enhancement
3733 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3734 Details : Update target/ChangeLog/which_patch .
3736 Severity : enhancement
3738 Description: jbd statistics
3739 Details : Port older jbd statistics patch for sles10
3744 Description: lvbo_init failed for resource with missing objects.
3745 Details : Fix returning error if we do stat for file with missing/corrupted
3746 objects and i_size set to all sum of size of all avaible objects.
3747 if we to truncate/write to missing object - it is recreated.
3752 Description: Oops in read and write path when failing to allocate lock.
3753 Details : Check if lock allocation failed and return error back.
3756 Frequency : When flocks are used.
3758 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3759 Details : Flock locks might destroy just granted lock if it could be merged
3760 with another existing flock, this is done in completion handler,
3761 so teach ldlm_cli_enquque_fini that this is a valid case for
3767 Description: reply_lock_interpret crash due to race with it and lock cancel.
3768 Details : Do not replay locks that are being cancelled. Do not reference
3769 locks by their address during replay, just by their handle.
3772 Frequency : only with deactivated OSTs
3774 Description: lstripe command fails for valid OST index
3775 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3776 instead of lov->desc.ld_active_tgt_count.
3778 Severity : enhancement
3780 Description: Build failure against Centos5 (RHEL5)
3781 Details : Use getpagesize() instead of PAGE_SIZE.
3783 --------------------------------------------------------------------------------
3785 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3787 * Support for kernels:
3788 2.6.5-7.283 (SLES 9),
3789 2.6.9-55.EL (RHEL 4),
3790 2.6.16.46-0.14 (SLES 10),
3791 2.6.18.8 vanilla (kernel.org)
3792 * Client support for unpatched kernels:
3793 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3794 2.6.16 - 2.6.22 vanilla (kernel.org)
3795 * Due to recently discovered recovery problems, we do not recommend
3796 using patchless RHEL 4 clients with this or any earlier release.
3797 * Recommended e2fsprogs version: 1.39.cfs8
3798 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3799 * Starting with this release, the ldiskfs backing filesystem required
3800 by Lustre is now in its own package, lustre-ldiskfs. This package
3801 should be installed. It is versioned separately from Lustre and
3802 may be released separately in future.
3807 Description: server LBUG when shutting down
3808 Details : Block umount forever until the mount refcount is zero rather
3809 than giving up after an arbitrary timeout.
3811 Severity : enhancement
3813 Description: add optional extra BUILD_VERSION info
3814 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3815 to override the lustre version.
3818 Frequency : 2.6.18 servers only
3820 Description: ll_kern_mount() doesn't release the module reference
3821 Details : The ldiskfs module reference count never drops down to 0
3822 because ll_kern_mount() doesn't release the module reference.
3827 Description: server LBUG when using old ost_num_threads parameter
3828 Details : Accept the old ost_num_threads parameter but warn that it
3829 is deprecated, and fix an off-by-one error that caused an LBUG.
3834 Description: Transient SCSI error results in persistent IO issue
3835 Details : iobuf->dr_error is not reinitialized to 0 between two
3839 Frequency : sometimes when underlying device returns I/O errors
3841 Description: OSTs not going read-only during write failures
3842 Details : OSTs are not remounted read-only when the journal commit threads
3843 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3844 instead of the ext3 wrappers.
3847 Frequency : only with 10000 clients or more
3849 Description: poor connect scaling with increasing client count
3850 Details : Don't run filter_grant_sanity_check for more than 100 exports
3851 to improve scaling for large numbers of clients.
3854 Frequency : SLES10 only
3856 Description: sanity-quota.sh quotacheck failed: rc = -22
3857 Details : Quotas cannot be enabled on SLES10.
3860 Frequency : liblustre clients only
3862 Description: getdirentries does not give error when run on compute nodes
3863 Details : getdirentries does not fail when the size specified as an argument
3864 is too small to contain at least one entry
3866 Severity : enhancement
3868 Description: Add LNET router traceability for debug purposes
3869 Details : If a checksum failure occurs with a router as part of the
3870 IO path, the NID of the last router that forwarded the bulk data
3871 is printed so it can be identified.
3876 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3877 Details : Due to a race condition, liblustre clients were occasionally
3878 evicted incorrectly.
3880 Severity : enhancement
3882 Description: lfs setstripe use optional parameters instead of postional
3885 Severity : enhancement
3887 Description: Nanosecond timestamp support for ldiskfs
3888 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3889 resolution timestamps. There is not yet support for this at
3890 the Lustre filesystem level.
3893 Frequency : during server recovery
3895 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3896 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3900 Frequency : nfs export on patchless client
3902 Description: connectathon hang when test nfs export over patchless client
3903 Details : Disconnected dentry cannot be found with lookup, so we do not need
3904 to unhash it or make it invalid
3907 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3908 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3909 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3914 Description: open req refcounting wrong on reconnect
3915 Details : If reconnect happened between getting open reply from server and
3916 call to mdc_set_replay_data in ll_file_open, we will schedule
3917 replay for unreferenced request that we are about to free.
3918 Subsequent close will crash in variety of ways.
3919 Check that request is still eligible for replay in
3920 mdc_set_replay_data().
3925 Description: disable writes to filesystem when reading health_check file
3926 Details : the default for reading the health_check proc file has changed
3927 to NOT do a journal transaction and write to disk, because this
3928 can cause reads of the /proc file to hang and block HA state
3929 checking on a healthy but otherwise heavily loaded system. It
3930 is possible to return to the previous behaviour during configure
3931 with --enable-health-write.
3933 Severity : enhancement
3935 Description: 64-bit inode version
3936 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3937 to the inode. This will be required for version-based recovery.
3942 Description: MDS fails to start if a duplicate client export is detected
3943 Details : in some rare cases it was possible for a client to connect to
3944 an MDS multiple times. Upon recovery the MDS would detect this
3945 and fail during startup. Handle this more gracefully.
3947 Severity : enhancement
3949 Description: Add -o localflock option to simulate old noflock behaviour.
3950 Details : This will achieve local-only flock/fcntl locks coherentness.
3955 Description: log_commit_thread vs filter_destroy race leads to crash
3956 Details : Take import reference before releasing llog record semaphore
3961 Description: Wrong request locking in request set processing
3962 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3963 imp_delayed_list, in this place should be used imp_lock.
3966 Frequency : when reconnecting
3968 Description: Grant leak when OSC reconnect to OST
3969 Details : When osc reconnect ost, OST(filter) should check whether it
3970 should grant more space to client by comparing fed_grant and
3971 cl_avail_grant, and return the granted space to client instead
3972 of "new granted" space, because client will call osc_init_grant
3973 to update the client grant space info.
3976 Frequency : when client reconnects to OST
3978 Description: Grant leak when OSC does a resend and replays bulk write
3979 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3980 bulk write request, because the grant info while be sync between
3981 OSC and OST when reconnect, and we should ignore the grant info
3982 these of resend/replay write req.
3987 Description: Grant space more than avaiable space sometimes.
3988 Details : When then OST is about to be full, if two bulk writing from
3989 different clients came to OST. Accord the avaliable space of the
3990 OST, the first req should be permitted, and the second one
3991 should be denied by ENOSPC. But if the seconde arrived before
3992 the first one is commited. The OST might wrongly permit second
3993 writing, which will cause grant space > avaiable space.
3996 Frequency : when client is evicted
3998 Description: Grant might be wrongly erased when osc is evicted by OST
3999 Details : when the import is evicted by server, it will fork another
4000 thread ptlrpc_invalidate_import_thread to invalidate the
4001 import, where the grant will be set to 0. While the original
4002 thread will update the grant it got when connecting. So if
4003 the former happened latter, the grant will be wrongly errased
4004 because of this race.
4009 Description: Checking Stale with correct fid
4010 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
4011 to check whether it is stale, because sometimes, we want the
4012 enqueue happened anyway, and op_data.fid2 will not be initialized.
4014 Severity : enhancement
4016 Description: update patchless client
4017 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
4020 Frequency : only with 2.4 kernel
4022 Description: random memory corruption
4023 Details : size of struct ll_inode_info is to big for union inode.u and this
4024 can be cause of random memory corruption.
4029 Description: Memory leak in recovery
4030 Details : Lov_mds_md was not free in an error handler in mds_create_object.
4031 It should also check obd_fail before fsfilt_start, otherwise if
4032 fsfilt_start return -EROFS,(failover mds during mds recovery).
4033 then the req will return with repmsg->transno = 0 and rc = EROFS.
4034 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
4035 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
4036 be freed no matter whether fsfilt_commit success or not.
4039 Frequency : only with large numbers of cores on a single node
4041 Description: Prevents from taking the superblock lock in llap_from_page for
4043 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
4044 for llap_from_page call in ll_removepage() prevents from taking
4045 the superblock lock for a soon died page.
4050 Description: Not check open intent error before release open handle
4051 Details : in some rare cases, the open intent error is not checked before
4052 release open handle, which may cause
4053 ASSERTION(open_req->rq_transno != 0), because it tries to release
4054 the failed open handle.
4059 Description: Set cat log bitmap only after create log success.
4060 Details : in some rare cases, the cat log bitmap is set too early. and it
4061 should be set only after create log success.
4065 Description: Accessing a block bevice can re-enable I/O when Lustre is
4066 tearing down a device.
4067 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
4072 Description: service threads may hog cpus when there are a lot of requests
4073 Details : Insert cond_resched to give other threads a chance to use some CPU
4078 Description: the cat log was not initialized in recovery
4079 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
4080 unlink log on mds will not be initialized until mds post
4081 recovery. And also in mds post recovery, the unlink log will
4082 initialization will be done asynchronausly, so there will be race
4083 between add unlink log and unlink log initialization.
4087 Description: brw_stats were being printed incorrectly
4088 Details : brw_stats were being printed as log2 but all of them were not
4089 recorded as log2. Also remove some code duplication arising from
4090 filter_tally_{read,write}.
4094 Frequency : rare, only in recovery.
4095 Description: ASSERTION(req->rq_type != LI_POISON) failed
4096 Details : imp_lock should be held while iterating over imp_sending_list for
4097 prevent destroy request after get timeout in ptlrpc_queue_wait.
4101 Description: replay-single.sh test 52 fails
4102 Details : A lock's skiplist need to be cleanup when it being unlinked
4103 from its resource list.
4108 Description: Short directio read returns full requested size rather than
4110 Details : Direct I/O operations should return actual amount of bytes
4111 transferred rather than requested size.
4113 Severity : enhancement
4115 Description: metadata RPC reduction (e.g. for rm performance)
4116 Details : decrease the amount of synchronous RPC between clients and servers
4117 by canceling conflicing lock before the operation on the client
4118 and packing thier handles into the main operation RPC to server.
4120 Severity : enhancement
4122 Description: add #ifdef HAVE_KERNEL_CONFIG_H
4123 Details : kernels from 2.6.19 not need include linux/config.h, but add
4124 include linux/autoconf.h in commpiler command line.
4126 Severity : enhancement
4128 Description: patchless client support for 2.6.22 kernel
4129 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
4130 removed. In this case we need drop using os depended interface to
4131 kmem_cache and use cfs_mem_cache API.
4136 Description: fix mal-formatted messages
4137 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
4141 Frequency : always in liblustre
4142 Description: wrong IS_ERR implementation in liblustre.h
4143 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
4148 Description: Correct condition for output debug message.
4149 Details : inode i_nlink equal zero is not enough for output message about
4150 disk corruption, i_ctime and i_mode should be also checked.
4154 Frequency : always in patchless client
4155 Description: add configure check for truncate_complete_page
4156 Details : improve checks for exported symbols. This allow run check without
4157 sources, but with Module.symvers shipped with kernel distribution.
4158 add check for truncate_complete_page used by patchless client.
4162 Description: sanity.sh test_77h fails with "test_77h file compare failed"
4163 Details : test_77h uses a file which was messed by other test case.
4167 Description: Not Check whether lov_tgts is NULL in some lov functions
4168 Details : Checking whether lov_tgts is NULL in some functions.
4172 Description: replace obdo_alloc() with OBDO_ALLOC macro
4173 Details : nothing special is done in obdo_alloc() function, and for
4174 debugging purpose, it needs to be replaced with macros.
4178 Description: bad return value and errno from fcntl call
4179 Details : In liblustre API, errno should be a negative value if error
4184 Description: ptlrpc_check_set() LBUG
4185 Details : In case of positive reply from server and failed client bulk
4186 callback after bulk transfer shouldn't LBUG, but process this
4187 request as erroneous.
4189 Severity : enhancement
4191 Description: VFS operations stats tool.
4192 Details : Tool which collects stats by tracking value written in pid,
4193 ppid, gid and uses llstat to generate output to plot graph using
4195 Updated lustre/utils/Makefile.am
4196 Added lustre/utils/ltrack_stats.c
4198 Severity : enhancement
4200 Description: 2.6.18 server support (lustre 1.6.1)
4201 Details : Support for 2.6.18 kernels on the server side.
4206 Description: ASSERTION(imp->imp_conn_current) failed
4207 Details : an assertion failure is hit if a client node boots and attempts to
4208 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
4211 Frequency : only for i686
4213 Description: 1.4.11 RC1 build fails for RHEL 4, i686
4214 Details : Fixed config variable for build.
4219 Description: Updated patchess for new RHEL4 kernel
4220 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
4221 Updated patch jbd-stats-2.6.9.patch
4222 Updated patch qsnet-rhel4-2.6.patch
4223 Updated patch quota-deadlock-on-pagelock-core.patch
4224 Updated patch vfs_intent-2.6-rhel4.patch
4225 Updated patch vfs_races-2.6-rhel4.patch
4226 Updated series file 2.6-rhel4-titech.series
4227 Updated series file 2.6-rhel4.series
4228 Updated kernel config files
4233 Description: lquota slave complains LBUG when reconnecting with mds
4235 Details : quota slave depends qctxt->lqc_import to send its quota request.
4236 This pointer will be invalid if mds did failover or broke its
4237 connect to osts, which leads to LBUG.
4240 Frequency : when qunit size is too small(less than 20M)
4242 Description: write is stopped by improper -EDQUOT
4243 Details : If the master is busy and qunit size is small enough(let's say 1M),
4244 the slave can not get quota from master on time, which will lead
4245 slave to trigger a -EQUOTA to client.
4250 Description: Deadlock during metadata tests
4251 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
4252 called with the per-dentry lock held.
4255 Frequency : SLES9 only
4257 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
4258 Details : set CONFIG_NR_CPUS to 128 instead of 8.
4260 Severity : enhancement
4262 Description: remove fs_prep_san_write operation and related patches
4263 Details : remove the ext3-san-jdike patches which are no longer useful.
4268 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4269 Details : a disk corruption can cause the mballoc code to assert on a
4270 double free or other extent corruptions. Handle these with
4271 ext3_error() instead of assertions.
4274 Frequency : only with mballoc3 code and deep extent trees
4276 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4277 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4278 can cause the filesystem to be remounted read-only.
4283 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4284 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4285 few external users of this ioctl number is lfsck's call to
4286 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4287 Add the old ioctl number to the handler so both old and new
4292 Description: parallel lock callbacks
4293 Details : Instead of sending blocking and completion callbacks as separated
4294 requests, adding them to a set and sending in parallel.
4298 Description: Disable most debugging by default
4299 Details : To improve performance, disable most logging (for debug purposes)
4300 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4301 default, and HA includes fewer messages.
4305 Frequency : only run on patchless client.
4306 Description: use do_facet on sanity.sh for test handling recoverables errors
4307 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4311 Description: hash tables for lists of nids, connections and uuids
4312 Details : Hash tables noticeably help when a lot of clients connect to a
4313 server, to faster identify duplicate connections or reconnects,
4314 also to faster find export to evict in manual eviction case.
4318 Description: Sometimes, when the server evict a client, and the client will
4319 not be evicted as soon as possible.
4320 Details : In enqueue req, the error was returned by intent, instead of
4321 rq_status which make ptlrpc layer not detect this error, and
4322 does not evict the client. So enqueue error should be returned
4326 Frequency : only at startup
4328 Description: Delay client connections to MDT until fist MDT->OST connect
4329 Details : If a client tried to create a new file before the MDT had
4330 connected to any OSTs, the create would return EIO. Now
4331 the client will simply block until the MDT connects to the
4332 first OST and the create can succeed.
4335 Frequency : only for SLES9
4337 Description: Routinely utilize latest Quadrics drivers in CFS releases
4338 Details : Update patch qsnet-suse-2.6.patch.
4341 Frequency : only for sles10
4343 Description: Update patches for SLES 10 SP1 kernel.
4344 Details : Update the patch vfs_intent-2.6-sles10.patch.
4349 Description: Routinely utilize latest Quadrics drivers in CFS releases
4350 Details : Update patch qsnet-rhel4-2.6.patch.
4353 Frequency : at statup only
4355 Description: mds_lov_synchronize race leads to various problems
4356 Details : simultaneous MDT->OST connections at startup can cause the
4357 sync to abort, leaving the OSC in a bad state.
4362 Description: Using wrong pointer in osc_brw_prep_request
4363 Details : Access to array[-1] can produce panic if kernel compiled with
4364 CONFIG_PAGE_ALLOC enabled
4366 Severity : enhancement
4368 Description: Async OSC create to avoid the blocking unnecessarily.
4369 Details : If a OST has no remain object, system will block on the creating
4370 when need to create a new object on this OST. Now, ways use
4371 pre-created objects when available, instead of blocking on an
4372 empty osc while others are not empty. If we must block, we block
4373 for the shortest possible period of time.
4377 Frequency : only in recovery
4378 Description: Mark OST as early accessible if his start SYNC.
4379 Details : osc_precreate return flag early accessible if oscc marked as
4380 OSCC_FLAG_SYNC_IN_PROGRESS.
4385 Description: Sometimes precreate code can triger create object on wrong ost
4386 Details : Wrong protected or not not restored variables aftre precreate loop
4387 can produce creation object on wrong ost.
4390 Frequency : oss recovery
4392 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4394 Details : llog_commit_thread_count should be synced between llog_commit
4395 start and cleanup, so new llog_commit thread should not be started
4396 when llog_commit threads being stopped to avoid accessing some
4399 Severity : enhancement
4401 Description: Add printing inode info into message about error in writepage.
4403 --------------------------------------------------------------------------------
4405 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4410 Frequency : on some architectures
4412 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4413 Details : Uninitialized flags sometimes cause configuration commands to
4417 Frequency : patchless clients only
4419 Description: missing __iget() symbol export
4420 Details : The __iget() symbol export is missing. To avoid the need for
4421 this on patchless clients the deathrow inode reaper is turned
4422 off, and we depend on the VM to clean up old inodes. This
4423 dependency was during via the fix for bug 12181.
4428 Description: sanity.sh fail: test_52b
4429 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4430 incorrect inode's flags to client.
4432 --------------------------------------------------------------------------------
4434 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4436 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4437 INTEROPERATE with older versions automatically. In many cases a
4438 special upgrade step is needed. Please read the
4439 user documentation before upgrading any part of a 1.4.x system.
4440 * WARNING: Lustre configuration and startup changes are required with
4441 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4443 * Support for kernels:
4444 2.4.21-47.0.1.EL (RHEL 3),
4445 2.6.5-7.283 (SLES 9),
4446 2.6.9-42.0.10.EL (RHEL 4),
4447 2.6.12.6 vanilla (kernel.org),
4448 2.6.16.27-0.9 (SLES10)
4449 * Client support for unpatched kernels:
4450 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4451 2.6.16 - 2.6.19 vanilla (kernel.org),
4452 2.6.9-42.0.8EL (RHEL 4)
4453 * Recommended e2fsprogs version: 1.39.cfs6
4454 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4457 Severity : enhancement
4459 Description: Async OSC create to avoid the blocking unnecessarily.
4460 Details : If a OST has no remain object, system will block on the creating
4461 when need to create a new object on this OST. Now, ways use
4462 pre-created objects when available, instead of blocking on an
4463 empty osc while others are not empty. If we must block, we block
4464 for the shortest possible period of time.
4466 Severity : enhancement
4468 Description: MountConf
4469 Details : Lustre configuration is now managed via mkfs and mount
4470 commands instead of lmc and lconf. New obd types (MGS, MGC)
4471 are added for dynamic configuration management. See
4472 https://mail.clusterfs.com/wikis/lustre/MountConf for
4475 Severity : enhancement
4477 Description: dynamic OST addition
4478 Details : OSTs can now be added to a live filesystem
4480 Severity : enhancement
4482 Description: startup order invariance
4483 Details : MDTs and OSTs can be started in any order. Clients only
4484 require the MDT to complete startup.
4486 Severity : enhancement
4488 Description: parallel, asynchronous orphan cleanup
4489 Details : orphan cleanup is now performed in separate threads for each
4490 OST, allowing parallel non-blocking operation.
4492 Severity : enhancement
4494 Description: optimized stripe assignment
4495 Details : stripe assignments are now made based on ost space available,
4496 ost previous usage, and OSS previous usage, in order to try
4497 to optimize storage space and networking resources.
4499 Severity : enhancement
4501 Description: Permanently set tunables
4502 Details : All writable /proc/fs/lustre tunables can now be permanently
4503 set on a per-server basis, at mkfs time or on a live
4506 Severity : enhancement
4508 Description: Lustre message v2
4509 Details : Add lustre message format v2.
4511 Severity : enhancement
4513 Description: client OST exclusion list
4514 Details : Clients can be started with a list of OSTs that should be
4515 declared "inactive" for known non-responsive OSTs.
4519 Description: ENOENT returned for valid filehandle during dbench.
4520 Details : Check if a directory has children when invalidating dentries
4521 associated with an inode during lock cancellation. This fixes
4522 an incorrect ENOENT sometimes seen for valid filehandles during
4523 testing with dbench.
4526 Frequency : SFS test only (otherwise harmless)
4528 Description: SPEC SFS validation failure on NFS v2 over lustre.
4529 Details : Changes the blocksize for regular files to be 2x RPC size,
4530 and not depend on stripe size.
4532 Severity : enhancement
4534 Description: fine-grained SMP locking inside DLM
4535 Details : Improve DLM performance on SMP systems by removing the single
4536 per-namespace lock and replace it with per-resource locks.
4538 Severity : enhancement
4540 Description: don't hold multiple extent locks at one time
4541 Details : To avoid client eviction during large writes, locks are not
4542 held on multiple stripes at one time or for very large writes.
4543 Otherwise, clients can block waiting for a lock on a failed OST
4544 while holding locks on other OSTs and be evicted.
4546 Severity : enhancement
4548 Description: Multiple MD RPCs in flight.
4549 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4550 To avoid overly-overloading MDT, introduce a limit on number of
4551 MDT RPCs in flight for a single client and add /proc controls
4552 to adjust this limit.
4554 Severity : enhancement
4556 Description: client read/write statistics
4557 Details : Add client read/write call usage stats for performance
4558 analysis of user processes.
4559 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4560 file access. extents_stats shows chunk size distribution.
4561 extents_stats_per_process show chunk size distribution per
4564 Severity : enhancement
4566 Description: per-client statistics on server
4567 Details : Add ldlm and operations statistics for each client in
4568 /proc/fs/lustre/mds|obdfilter/*/exports/
4570 Severity : enhancement
4572 Description: improved MDT statistics
4573 Details : Add detailed MDT operations statistics in
4574 /proc/fs/lustre/mds/*/stats
4576 Severity : enhancement
4578 Description: VFS operations stats
4579 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4580 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4585 Description: Fix client-side osc byte counters
4586 Details : The osc read/write byte counters in
4587 /proc/fs/lustre/osc/*/stats are now working
4590 Frequency : always as root on SLES
4592 Description: Failure of copying files with lustre special EAs.
4593 Details : Client side always return success for setxattr call for lustre
4594 special xattr (currently only "trusted.lov").
4599 Description: Refcount LNET uuids
4600 Details : The global LNET uuid list grew linearly with every startup;
4601 refcount repeated list entries instead of always adding to
4604 Severity : enhancement
4606 Description: Dynamic service threads
4607 Details : Within a small range, start extra service threads
4608 automatically when the request queue builds up.
4611 Frequency : mixed-endian client/server environments
4613 Description: mixed-endian crashes
4614 Details : The new msg_v2 system had some failures in mixed-endian
4617 Severity : enhancement
4619 Description: Easy OST removal
4620 Details : OSTs can be permanently deactivated with e.g. 'lctl
4621 conf_param lustre-OST0001.osc.active=0'
4623 Severity : enhancement
4625 Description: MGS proc entries
4626 Details : Added basic proc entries for the MGS showing what filesystems
4629 Severity : enhancement
4631 Description: provide MGS failover
4632 Details : Added config lock reacquisition after MGS server failover.
4634 Severity : enhancement
4636 Description: add Linux 2.4 support
4637 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4641 Description: a large application tries to do I/O to the same resource and dies
4642 in the middle of it.
4643 Details : Check the req->rq_arrival time after the call to
4644 ost_brw_lock_get(), but before we do anything about
4645 processing it & sending the BULK transfer request. This
4646 should help move old stale pending locks off the queue as
4647 quickly as obd_timeout.
4650 Frequency : when an incorrect nid is specified during startup
4652 Description: ptlrpc connect to non-existant node causes kernel crash
4653 Details : LNET can't be re-entered from an event callback, which
4654 happened when we expire a message after the export has been
4655 cleaned up. Instead, hand the zombie cleanup off to another
4658 Severity : enhancement
4660 Description: plain/inodebits lock performance improvement
4661 Details : Grouping plain/inodebits in granted list by their request modes
4662 and bits policy, thus improving the performance of search through
4666 Frequency : only if OST filesystem is corrupted
4668 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4669 Details : for a short time RPCs with bulk IO are in the replay list,
4670 but replay of bulk IOs is unimplemented. If the OST filesystem
4671 is corrupted due to disk cache incoherency and then replay is
4672 started it is possible to trip an assertion. Avoid putting
4673 committed RPCs into the replay list at all to avoid this issue.
4676 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4678 Description: System hang on startup
4679 Details : This bug allowed the liblustre (e.g. catamount) client to
4680 return to the app before handling all startup RPCs. This
4681 could leave the node unresponsive to lustre network traffic
4682 and manifested as a server ptllnd timeout.
4684 Severity : enhancement
4686 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4687 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4688 boolean to control whether to print peer debug info when a
4689 client's RPC times out.
4692 Frequency : only for kernels with patches from Lustre below 1.4.3
4694 Description: Remove old rdonly API
4695 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4698 Frequency : only for devices with external journals
4700 Description: Set external device read-only also
4701 Details : During a commanded failover stop, we set the disk device
4702 read-only while the server shuts down. We now also set any
4703 external journal device read-only at the same time.
4706 Frequency : when upgrading from 1.4 while trying to change parameters
4708 Description: The wrong (new) MDC name was used when setting parameters for
4709 upgraded MDT's. Also allows changing of OSC (and MDC)
4710 parameters if --writeconf is specified at tunefs upgrade time.
4713 Frequency : when setting specific ost indicies
4715 Description: QOS code breaks on skipped indicies
4716 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4717 created with --index need not be sequential.
4719 Severity : enhancement
4721 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4722 Details : The uninit_groups feature works in conjunction with the kernel
4723 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4724 up the pass1 processing of e2fsck. This is a read-only feature
4725 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4726 mount filesystems that have had this feature enabled.
4728 Severity : enhancement
4730 Description: Improve multi-block allocation algorithm to avoid fragmentation
4731 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4732 allocation locality and avoid filesystem fragmentation.
4734 ------------------------------------------------------------------------------
4736 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4738 * Support for kernels:
4739 2.4.21-47.0.1.EL (RHEL 3)
4740 2.6.5-7.283 (SLES 9)
4741 2.6.9-42.0.10.EL (RHEL 4)
4742 2.6.12.6 vanilla (kernel.org)
4743 2.6.16.27-0.9 (SLES 10)
4744 * Recommended e2fsprogs version: 1.39.cfs5
4746 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4750 Frequency : occasional, depends on client load and configuration
4751 Bugzilla : 12181, 12203
4752 Description: data loss for recently-modified files
4754 Details : In some cases it is possible that recently written or created
4755 files may not be written to disk in a timely manner (this should
4756 normally be within 30s unless client IO load is very high).
4757 The problem appears as zero-length files or files that are a
4758 multiple of 1MB in size after a client crash or client eviction
4759 that are missing data at the end of the file.
4761 This problem is more likely to be hit on clients where files are
4762 repeatedly created and unlinked in the same directory, clients
4763 have a large amount of RAM, have many CPUs, the filesystem has
4764 many OSTs, the clients are rebooted frequently, and/or the files
4765 are not accessed by other nodes after being written.
4767 The presence of the problem can be detected by looking at
4768 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4769 smaller than the second (nr_unused) then dirty files will not
4770 be flushed automatically to disk. "sync; sleep 10" should be
4771 run several times on the node before unmounting it to update
4772 Lustre (this is also safe to run on nodes without this problem).
4774 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4775 that can cause this same problem, so customers using that kernel
4776 also need to update the kernel in addition to Lustre. In order
4777 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4779 It is normal that files written just before a client crash (less
4780 than 30s) may not yet have been flushed to disk, even for local
4784 Frequency : frequent on thin XT3 nodes
4786 Description: UUID collision on thin XT3 Linux nodes
4787 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4788 randomly, since we relied on an insufficiently-seeded PRNG.
4793 Description: OSS hangs after "All ost request buffers busy"
4794 Details : A deadlock between quota and journal operations caused OSS
4795 hangs after printing "All ost request buffers busy."
4798 Frequency : always on liblustre builds
4800 Description: Cleanup compiler warnings on liblustre
4803 Frequency : always on liblustre builds on XT3
4805 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4809 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4810 > 32000 subdirectories
4811 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4812 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4813 subdirectory count crosses 32000. This will aid e2fsck to
4814 correctly handle more than 32000 subdirectories.
4817 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4819 Description: System hang on startup
4820 Details : This bug allowed the liblustre (e.g. catamount) client to
4821 return to the app before handling all startup RPCs. This
4822 could leave the node unresponsive to lustre network traffic
4823 and manifested as a server ptllnd timeout.
4825 Severity : enhancement
4827 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4828 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4829 boolean to control whether to print peer debug info when a
4830 client's RPC times out.
4835 Description: make O_SYNC working on 2.6 kernels
4836 Details : 2.6 kernels use different method for mark pages for write,
4837 so need add a code to lustre for O_SYNC work.
4842 Description: Failure to close file and release space on NFS
4843 Details : Put inode details into lock acquired in ll_intent_file_open.
4844 Use mdc_intent_lock in ll_intent_open to properly
4845 detect all kind of errors unhandled by mdc_enqueue.
4850 Description: proc file read during shutdown sometimes raced obd removal,
4852 Details : Add lock to prevent obd access after proc file removal.
4855 Frequency : Only for files larger than 4GB on 32-bit clients.
4857 Description: improperly doing page alignment of locks
4858 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4859 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4864 Description: RPCs being resent when they shouldn't be
4865 Details : Some RPCs that should not be resent are being resent. This
4866 can cause inconsistencies in the RPC state machine. Do not
4867 resend such requests.
4870 Frequency : rare, only with NFS export
4872 Description: Crash on NFS re-export node
4873 Details : under very unusual load conditions an assertion is hit in
4874 ll_intent_file_open()
4877 Frequency : only if OST filesystem is corrupted
4879 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4880 Details : for a short time RPCs with bulk IO are in the replay list,
4881 but replay of bulk IOs is unimplemented. If the OST filesystem
4882 is corrupted due to disk cache incoherency and then replay is
4883 started it is possible to trip an assertion. Avoid putting
4884 committed RPCs into the replay list at all to avoid this issue.
4889 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4890 Details : Large single O_DIRECT read and write calls can fail to allocate
4891 a sufficiently large buffer to process the request. In case of
4892 allocation failure the allocation is retried with a smaller
4893 buffer and broken into smaller requests.
4895 Severity : enhancement
4897 Description: Add -o localflock option to simulate old noflock behaviour.
4898 Details : This will achieve local-only flock/fcntl locks coherentness.
4903 Description: versioning check is incomplete
4904 Details : Checking the version difference of client vs. server, report
4905 error if the gap is too big.
4910 Description: add support PG_writeback bit
4911 Details : add support for PG_writeback bit for Lustre, for more carefull
4912 work with page cache in 2.6 kernel. This also fix some deadlocks
4913 and remove hack for work O_SYNC with 2.6 kernel.
4915 Severity : enhancement
4917 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4918 Details : The uninit_groups feature works in conjunction with the kernel
4919 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4920 up the pass1 processing of e2fsck. This is a read-only feature
4921 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4922 mount filesystems that have had this feature enabled.
4924 Severity : enhancement
4926 Description: Improve multi-block allocation algorithm to avoid fragmentation
4927 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4928 allocation locality and avoid filesystem fragmentation.
4930 ------------------------------------------------------------------------------
4932 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4934 * Support for kernels:
4935 2.6.9-42.0.3.EL (RHEL 4)
4936 2.6.5-7.276 (SLES 9)
4937 2.4.21-47.0.1.EL (RHEL 3)
4938 2.6.12.6 vanilla (kernel.org)
4939 2.6.16.21-0.8 (SLES10)
4940 * Recommended e2fsprogs version: 1.39.cfs2-0
4942 * The backwards-compatible /proc/sys/portals symlink has been removed
4943 in this release. Before upgrading, please ensure that you change
4944 any configuration scripts or /etc/sysctl.conf files that access
4945 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4946 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4947 in advance of the upgrade on any system running Lustre 1.4.6 or
4948 newer, since /proc/sys/lnet was added in that version.
4949 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4953 Frequency : only when quota is used
4955 Description: avoid scanning export list for quota master
4956 Details : Change the algorithms to avoid scanning export list in order
4957 to improve the efficiency.
4960 Frequency : MDS failover only, very rarely
4962 Description: "went back in time" messages on mds failover
4963 Details : The greatest transno may be lost when the current operation
4964 finishes with an error (transno==0) and the client's last_rcvd
4965 record is over-written. Save the greatest transno in the
4966 mds_last_transno for this case.
4969 Frequency : always for specific kernels and striping counts
4971 Description: client may get "Matching packet too big" without ACL support
4972 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4973 when trying to access files in certain configurations. The
4974 clients should in fact be denied when mounting because they do
4975 not understand ACLs.
4978 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4980 Description: many clients connecting with IO in progress causes connect timeouts
4981 Details : Avoid synchronous journal commits to avoid delays caused by many
4982 clients connecting/disconnecting when bulk IO is in progress.
4983 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4984 OST_IO_PORTAL to avoid delays behind potentially many pending
4988 Frequency : occasionally with multiple writers to a single file
4990 Description: shared writes to file may result in wrong size reported by stat()
4991 Details : Allow growing of kms when extent lock is cancelled
4994 Frequency : always with random mmap IO to multi-striped file
4996 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4997 Details : Only if the hole is at the end of OST object so that kms is too
4998 small. Fix is to increase kms accordingly in ll_nopage.
5001 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
5003 Description: writes to a missing object would leak memory on the OST
5004 Details : If there is an inconsistency between the MDS and OST filesystems,
5005 such that the MDS references an object that doesn't exist, writes
5006 to that object will leak memory due to incorrect cleanup in the
5007 error handling path, eventually running out of memory on the OST.
5012 Description: Creating too long symlink causes lustre errors
5013 Details : Check symlink and name lengths before sending requests to MDS.
5016 Frequency : only if flock is enabled (not on by default)
5018 Description: posix locks not released on fd closure on 2.6.9+
5019 Details : We failed to add posix locks to list of inode locks on 2.6.9+
5020 kernels, this caused such locks not to be released on fd close and
5021 then assertions on fs unmount about still used locks.
5024 Frequency : MDS failover only, very rarely
5026 Description: clients may get ASSERTION(granted_lock != NULL)
5027 Details : When request was taking a long time, and a client was resending
5028 a getattr by name lock request. The were multiple lock requests
5029 with the same client lock handle and
5030 mds_getattr_name->fixup_handle_for_resent_request found one of the
5031 lock handles but later failed with ASSERTION(granted_lock != NULL).
5036 Description: handle->h_buffer_credits > 0, assertion failure
5037 Details : h_buffer_credits is zero after truncate, causing assertion
5038 failure. This patch extends the transaction or creates a new
5042 Frequency : NFS re-export or patchless client
5043 Bugzilla : 11179, 10796
5044 Description: Crash on NFS re-export node (__d_move)
5045 Details : We do not want to hash the dentry if we don't have a lock.
5046 But if this dentry is later used in d_move, we'd hit uninitialised
5047 list head d_hash, so we just do this to init d_hash field but
5048 leave dentry unhashed.
5051 Frequency : NFS re-export or patchless client
5053 Description: NFS exports has problem with symbolic link
5054 Details : lustre client didn't properly install dentry when re-exported
5055 to NFS or running patchless client.
5058 Frequency : NFS re-export or patchless client
5060 Description: Various nfs/patchless fixes.
5061 Details : fixes reuse disconected alias for lookup process - this fixes
5062 warning "find_exported_dentry: npd != pd",
5063 fix permission error with open files at nfs.
5064 fix apply umask when do revalidate.
5067 Frequency : occasional
5069 Description: Crash on NFS re-export node
5070 Details : calling clear_page() on the wrong pointer triggered oops in
5071 generic_mapping_read().
5074 Frequency : rarely, using O_DIRECT IO
5076 Description: unaligned directio crashes client with LASSERT
5077 Details : check for unaligned buffers before trying any requests.
5080 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
5082 Description: stale data returned from RAID cache
5083 Details : If only a small amount of IO is done to the RAID device before
5084 reading it again it is possible to get stale data from the RAID
5085 cache instead of reading it from disk.
5088 Frequency : always for sles10 kernel
5090 Description: sles10 support
5091 Details : ll_follow_link: compile fixes and using of nd_set_link
5092 under newer kernels.
5095 Frequency : depends on arch, kernel and compiler version, always on sles10
5098 Description: recursive or deep enough symlinks cause stack overflow
5099 Details : getting rid of large stack-allocated variable in
5103 Frequency : depends on hardware
5105 Description: lustre write performance loss in the SLES10 kernel
5106 Details : the performance loss is caused by using of write barriers in the
5107 ext3 code. The SLES10 kernel turns barrier support on by
5108 default. The fix is to undo that change for ldiskfs.
5110 ------------------------------------------------------------------------------
5112 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
5114 * Support for kernels:
5115 2.6.9-42.0.3EL (RHEL 4)
5116 2.6.5-7.276 (SLES 9)
5117 2.4.21-47.0.1.EL (RHEL 3)
5118 2.6.12.6 vanilla (kernel.org)
5122 Frequency : quota enabled and large files being deleted
5124 Description: releasing more than 4GB of quota at once hangs OST
5125 Details : If a user deletes more than 4GB of files on a single OST it
5126 will cause the OST to spin in an infinite loop. Release
5127 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
5132 Description: statfs data retrieved from /proc may be stale or zero
5133 Details : When reading per-device statfs data from /proc, in the
5134 {kbytes,files}_{total,free,avail} files, it may appear
5135 as zero or be out of date.
5138 Frequency : systems with MD RAID1 external journal devices
5140 Description: lconf's call to blkid is confused by RAID1 journal devices
5141 Details : Use the "blkid -l" flag to locate the MD RAID device instead
5142 of returning all block devices that match the journal UUID.
5145 Frequency : always, for aggregate stripe size over 4GB
5147 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
5148 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
5149 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
5152 Frequency : always if "lfs find" used on a local file/directory
5154 Description: "lfs find" segfaults if used on a local file/directory
5155 Details : The case where a directory component was not specified wasn't
5156 handled correctly. Handle this properly.
5159 Frequency : always on ppc64
5161 Description: the write to an ext3 filesystem mounted with mballoc got stuck
5162 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
5163 perform endianness conversion.
5166 Frequency : rarely (truncate to non-zero file size after write under load)
5167 Bugzilla : 10730, 10687
5168 Description: Files padded with zeros to next 4K multiple
5169 Details : With filesystems mounted using the "extents" option (2.6 kernels)
5170 it is possible that files that are truncated to a non-zero size
5171 immediately after being written are filled with zero bytes beyond
5172 the truncated size. No file data is lost.
5174 Severity : enhancement
5176 Description: Allow recovery/failover for liblustre clients.
5177 Details : liblustre clients were unaware of failover configurations until
5180 Severity : enhancement
5182 Description: user file locks should fail when not mounting with flock option
5183 Details : Set up an error-returning stub in ll_file_operations.lock field
5184 to prevent incorrect behaviour when client is mounted without
5185 flock option. Also, set up properly f_op->flock field for
5189 Frequency : always on ia64
5191 Description: "lfs df" loops on printing out MDS statfs information
5192 Details : The obd_ioctl_data was not initialized and in some systems
5193 this caused a failure during the ioctl that did not return
5194 an error. Initialize the struct and return an error on failure.
5197 Frequency : SLES 9 only
5199 Description: Error of copying files with lustre special EAs as root
5200 Details : Client side always return success for setxattr call for lustre
5201 special xattr (currently only "trusted.lov").
5204 Frequency : rarely on clusters with both ia64+i386 clients
5206 Description: ia64+i686 clients doing shared IO on the same file may LBUG
5207 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
5208 clients are doing concurrent writes to the same file it is
5209 possible that the ia64 clients may LASSERT because the OST
5210 extent locks are not PAGE_SIZE aligned. Ensure that grown
5211 locks are always aligned on the request boundary.
5214 Frequency : specific use, occasional
5216 Description: Overwriting in use executable truncates on-disk binary image
5217 Details : If one node attempts to overwrite an executable in use by
5218 another node, we now correctly return ETXTBSY instead of
5219 truncating the file.
5221 Severity : enhancement
5223 Description: Async OSC create to avoid the blocking unnecessarily.
5224 Details : If a OST has no remain object, system will block on the creating
5225 when need to create a new object on this OST. Now, ways use
5226 pre-created objects when available, instead of blocking on an
5227 empty osc while others are not empty. If we must block, we block
5228 for the shortest possible period of time.
5233 Description: chmod on Lustre root is propagated to other clients
5234 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
5235 invalid by the follow_mount time.
5238 Frequency : liblustre clients only
5240 Description: Race in 'instant cancel' lock handling could lead to such locks
5241 never to be granted in case of SMP MDS
5242 Details : Do not destroy not yet granted but cbpending locks in
5246 Frequency : replay/resend of open
5248 Description: non null lock assetion failure in mds_intent_policy
5249 Details : Trying to replay/resend lockless open requests resulted in
5250 mds_open() returning 0 with no lock. Now it sets a flag if
5251 a lock is going to be returned.
5253 Severity : enhancement
5255 Description: Checksum enhancements
5256 Details : New checksum enhancements allow for resending RPCs that failed
5259 Severity : enhancement
5261 Description: Tunables on number of dirty pages in cacche
5262 Details : Allow to set limit on number of dirty pages cached.
5267 Description: client crash on unmount - lock still has references
5268 Details : In some error handling cases it was possible to leak a lock
5269 reference on a client while accessing a file. This was not
5270 harmful to the client during operation, but would cause the
5271 client to crash when the filesystem is unmounted.
5274 Frequency : specific case, rare
5276 Description: ETXTBSY on mds though file not in use by client
5277 Details : ETXTBSY is no longer incorrectly returned when attempting to
5278 chmod or chown a directory that the user previously tried to
5279 execute or a currently-executing binary.
5282 Frequency : extremely rare except on liblustre-based clients
5284 Description: Lustre space not freed when files are deleted
5285 Details : Clean up open-unlinked files after client eviction. Previously
5286 the unlink was skipped and the files remained as orphans.
5291 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5292 Details : In some cases it was possible to send a blocking callback to
5293 a client doing a glimpse, even though that client didn't get
5294 a lock granted. When the glimpse lock is cancelled on the OST
5295 the freed lock is left on the waiting list and corrupted the list.
5298 Frequency : all core dumps
5300 Description: Broke core dumps to lustre
5301 Details : Negative dentry may be unhashed if parent does not have UPDATE
5302 lock, but some callers, e.g. do_coredump, expect dentry to be
5303 hashed after successful create, hash it in ll_create_it.
5305 ------------------------------------------------------------------------------
5307 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5309 * Support for kernels:
5310 2.6.9-42.0.2.EL (RHEL 4)
5311 2.6.5-7.276 (SLES 9)
5312 2.4.21-40.EL (RHEL 3)
5313 2.6.12.6 vanilla (kernel.org)
5317 Frequency : always on RHEL 3
5319 Description: Number of open files grows over time
5320 Details : The number of open files grows over time, whether or not
5321 Lustre is started. This was due to a filp leak introduced
5322 by one of our kernel patches.
5324 ------------------------------------------------------------------------------
5326 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5328 * Support for kernels:
5329 2.6.9-42.EL (RHEL 4)
5330 2.6.5-7.267 (SLES 9)
5331 2.4.21-40.EL (RHEL 3)
5332 2.6.12.6 vanilla (kernel.org)
5337 Bugzilla : 5719, 9635, 9792, 9684
5338 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5339 Details : If a server is under heavy load and cannot reply to new
5340 connection requests before the client resends the (re)connect,
5341 the connection handling code can behave badly if two service
5342 threads are concurrently handing separate (re)connections from
5343 the same client. Add better locking to the connection handling
5344 code, and ensure that only a single connection will be processed
5345 for a given client UUID, even if the lock is dropped.
5347 Severity : enhancement
5349 Description: add TCP zero-copy support to kernel
5350 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5351 sends if the hardware supports scatter-gather and checksumming.
5352 This allows socklnd to do client-write and server-read more
5353 efficiently and reduce CPU utilization from skbuf copying.
5356 Frequency : only if NFS exporting from client
5358 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5359 Details : If setting attributes on a file created under NFS that had
5360 never been opened it would be possible to oops the client
5361 if the file had no objects.
5365 Bugzilla : 9326, 10402, 10897
5366 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5367 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5368 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5369 before ptlrpcd_wake() was called and the request was freed.
5372 Frequency : always for liblustre
5374 Description: liblustre client does MDS+OSTs setattr RPC for each write
5375 Details : When doing a write from a liblustre client, the client
5376 incorrectly issued an RPC to the MDS and each OST the file was
5377 striped over in order to update the timestamps. When writing
5378 with small chunks and many clients this could overwhelm the MDS
5379 with RPCs. In all cases it would slow down the write because
5380 these RPCs are unnecessary.
5382 Severity : enhancement
5384 Description: allow number of MDS service threads to be changed at module load
5385 Details : It is now possible to change the number of MDS service threads
5386 running. Adding "options mds mds_num_threads={N}" to the MDS's
5387 /etc/modprobe.conf will set the number of threads for the next
5388 time Lustre is restarted (assuming the "mds" module is also
5389 reloaded at that time). The default number of threads will
5390 stay the same, 32 for most systems.
5395 Description: OST crash if filesystem is unformatted or corrupt
5396 Details : If an OST is started on a device that has never been formatted
5397 or if the filesystem is corrupt and cannot even mount then the
5398 error handling cleanup routines would dereference a NULL pointer.
5403 Description: NULL pointer deref in llap_from_page.
5404 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5405 incorrect) mapping if the page was truncated/reclaimed while it was
5406 searched for. Check for this condition and skip such pages when
5407 doing readahead. Introduce extra check to llap_from_page() to
5408 verify page->mapping->host is non-NULL (so page is not anonymous).
5411 Frequency : Sometimes when using sys_sendfile
5413 Description: "page not covered by a lock" warnings from ll_readpage
5414 Details : sendfile called ll_readpage without right page locks present.
5415 Now we introduced ll_file_sendfile that does necessary locking
5416 around call to generic_file_sendfile() much like we do in
5420 Frequency : with certain MDS communication failures at client mount time
5422 Description: NULL pointer deref after failed client mount
5423 Details : a client connection request may delayed by the network layer
5424 and not be sent until after the PTLRPC layer has timed out the
5425 request. If the client fails the mount immediately it will try
5426 to clean up before the network times out the request. Add a
5427 reference from the request import to the obd device and delay
5428 the cleanup until the network drops the request.
5431 Frequency : occasionally during client (re)connect
5433 Description: assertion failure during client (re)connect
5434 Details : processing a client connection request may be delayed by the
5435 client or server longer than the client connect timeout. This
5436 causes the client to resend the connection request. If the
5437 original connection request is replied in this interval, the
5438 client may trip an assertion failure in ptlrpc_connect_interpret()
5439 which thought it would be the only running connect process.
5442 Frequency : only with obd_echo servers and clients that are rebooted
5444 Description: kernel BUG accessing uninitialized data structure
5445 Details : When running an obd_echo server it did not start the ping_evictor
5446 thread, and when a client was evicted an uninitialized data
5447 structure was accessed. Start the ping_evictor in the RPC
5448 service startup instead of the OBD startup.
5450 Severity : enhancement
5451 Bugzilla : 10193 (patchless)
5452 Description: Remove dependency on various unexported kernel interfaces.
5453 Details : No longer need reparent_to_init, exit_mm, exit_files,
5454 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5457 Frequency : rare (only users of deprecated and unsupported LDAP config)
5459 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5460 Details : LDAP apparently contains 'lustreName' attributes instead of
5461 'name'. A simple remapping of the name is sufficient.
5464 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5466 Description: waiting_locks_callback trips kernel BUG if client is evicted
5467 Details : Running with the dump_on_timeout debug flag turned on makes
5468 it possible that the waiting_locks_callback() can try to dump
5469 the Lustre kernel debug logs from an interrupt handler. Defer
5470 this log dumping to the expired_lock_main() thread.
5472 Severity : enhancement
5474 Description: Support NFS exporting on 2.6 kernels.
5475 Details : Implement non-rawops metadata methods for NFS server to use without
5476 changing NFS server code.
5479 Frequency : very rare (synthetic metadata workload only)
5481 Description: two racing renames might cause an MDS thread to deadlock
5482 Details : Running the "racer" program may cause one MDS thread to rename
5483 a file from being the source of a rename to being the target of
5484 a rename at exactly the same time that another thread is doing
5485 so, and the second thread has already enqueued these locks after
5486 doing a lookup of the target and is trying to relock them in
5487 order. Ensure that we don't try to re-lock the same resource.
5490 Frequency : only very large systems with liblustre clients
5492 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5493 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5494 all OSTs in parallel. This allows the eviction of stale liblustre
5495 clients to proceed much faster than if they were done in series,
5496 and also offers similar improvements for other set_info RPCs.
5501 Description: excessive CPU usage during initial read phase on client
5502 Details : During the initial read phase on a client, it would agressively
5503 retry readahead on the file, consuming too much CPU and impacting
5504 performance (since 1.4.5.8). Improve the readahead algorithm
5505 to avoid this, and also improve some other common cases (read
5506 of small files in particular, where "small" is files smaller than
5507 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5512 Description: MDS crash when receiving packet with unknown intent.
5513 Details : Do not LBUG in unknown intent case, just return -EFAULT
5515 Severity : enhancement
5516 Bugzilla : 9293, 9385
5517 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5518 Details : Do not serialize getattr (non-intent version) and statfs.
5521 Frequency : occasional, when OST network is overloaded/intermittent
5523 Description: client evicted by OST after bulk IO timeout
5524 Details : If a client sends a bulk IO request (read or write) the OST
5525 may evict the client if it is unresposive to its data GET/PUT
5526 request. This is incorrect if the network is overloaded (takes
5527 too long to transfer the RPC data) or dropped the OST GET/PUT
5528 request. There is no need to evict the client at all, since
5529 the pinger and/or lock callbacks will handle this, and the
5530 client can restart the bulk request.
5533 Frequency : Always when mmapping file with no objects
5535 Description: client crashes when mmapping file with no objects
5536 Details : Check that we actually have objects in a file before doing any
5537 operations on objects in ll_vm_open, ll_vm_close and
5543 Description: Request leak when working with deleted CWD
5544 Details : Introduce advanced request refcount tracking for requests
5545 referenced from lustre intent.
5547 Severity : Enhancement
5549 Description: Cache open file handles on client.
5550 Details : MDS now will return special lock along with openhandle, if
5551 requested and client is allowed to hold openhandle, even if unused,
5552 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5553 closing files for every read/write openration.
5555 Severity : Enhancement
5557 Description: Cache open negative dentries on client when possible.
5558 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5559 negative dentries on lock revocation.
5564 Description: Remounting a client read-only wasn't possible with a zconf mount
5565 Details : It wasn't possible to remount a client read-only with llmount.
5567 Severity : enhancement
5568 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5569 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5570 MPICH is included to add Lustre-specific ADIO interfaces.
5571 This is based closely on the UFS ADIO layer and only differs
5572 in file creation, in order to allow the OST striping to be set.
5573 This is user-contributed code and not supported by CFS.
5578 Description: extended inode attributes (immutable, append-only) work improperly
5579 when 2.4 and 2.6 kernels are used on client/server or vice versa
5580 Details : Introduce kernel-independent values for these flags.
5582 Severity : enhancement
5585 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5586 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5587 granularity. Now, Fractional MB granularity is be supported,
5588 this is very useful for low memory system.
5590 Severity : enhancement
5592 Description: Getattr by fid
5593 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5594 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5595 have locks and avoids getattr rpc requests for it on every stat.
5598 Frequency : Always, for filesystems larger than 2TB
5600 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5601 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5602 filesystems larger than 2TB to be mounted with mballoc enabled.
5605 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5607 Description: filesystem corruption for non-standard kernels and very large OSTs
5608 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5609 filesystems larger than 2TB is mounted then the kernel will
5610 silently corrupt the start of the filesystem. CONFIG_LBD is
5611 enabled for all CFS-supported kernels, but the possibility of
5612 this happening with a modified kernel config exists.
5614 Severity : enhancement
5616 Description: add client O_DIRECT support for 2.6 kernels
5617 Details : It is now possible to do O_DIRECT reads and writes to files
5618 in the Lustre client mountpoint on 2.6 kernel clients.
5620 Severity : enhancement
5622 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5623 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5624 parallel, not waiting for response from every OST before sending
5625 a rpc to the next OST.
5630 Description: setattr vs write race when updating file timestamps
5631 Details : Client processes that update a file timestamp into the past
5632 right after writing to the file (e.g. tar) it is possible that
5633 the updated file modification time can be reset to the current
5634 time due to a race between processing the setattr and write RPC.
5636 Severity : enhancement
5638 Description: Bring 'lfs find' closer in line with regular Linux find.
5639 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5640 -print0 options and obtains all the needed info through the lustre
5643 Severity : enhancement
5645 Description: support up to 1024 configured devices on one node
5646 Details : change obd_dev array from statically allocated to dynamically
5647 allocated structs as they are first used to reduce memory usage
5652 Description: Flush dirty partially truncated pages during truncate
5653 Details : Immediatelly flush partially truncated pages in filter_setattr,
5654 this way we completely avoid having any pages in page cache on OST
5655 and can retire ugly workarounds during writes to flush such pages.
5660 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5661 Details : protect inode from truncation within vfs_unlink() context
5662 just take a reference before calling vfs_unlink() and release it
5663 when parent's i_sem is free.
5666 Frequency : always, if extents are used on OSTs
5668 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5669 Details : OSTs using the extents format would not zero the high 16 bits of
5670 the index physical block number. This is not a problem for any
5671 OST filesystems smaller than 16TB, and no kernels support ext3
5672 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5673 new/modified files) and can be fixed for existing filesystems
5674 with e2fsprogs-1.39-cfs1.
5679 Description: import connection selection may be incorrect if timer wraps
5680 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5681 import connections to be ignored if the 32-bit jiffies counter
5682 wraps. Use a 64-bit jiffies counter.
5685 Frequency : during server recovery
5687 Description: crash after server is denying duplicate export
5688 Details : If clients are resending connect requests to the server, the
5689 server refuses to allow a client to connect multiple times.
5690 Fixed a bug in the handling of this case.
5693 Frequency : very large clusters immediately after boot
5695 Description: LNET request buffers exhausted under heavy short-term load
5696 Details : If a large number of client requests are generated on a service
5697 that has previously never seen so many requests it is possible
5698 that the request buffer growth cannot keep up with the spike in
5699 demand. Instead of dropping incoming requests, they are held in
5700 the LND until the RPC service can accept more requests.
5703 Frequency : Sometimes during replay
5705 Description: Assertion failure in ll_local_open after replay.
5706 Details : If replay happened on an open request reply before we were able
5707 to set replay handler, reply will become not swabbed tripping the
5708 assertion in ll_local_open. Now we set the handler right after
5709 recognising of open request
5712 Frequency : very rare
5714 Description: kernel reports "badness in vsnprintf"
5715 Details : Reading from the "recovery_status" /proc file in small chunks
5716 may cause a negative length in lprocfs_obd_rd_recovery_status()
5717 call to vsnprintf() (which is otherwise harmless). Exit early
5718 if there is no more space in the output buffer.
5720 Severity : enhancement
5722 Description: clear OBD RPC statistics by writing to them
5723 Details : It is now possible to clear the OBD RPC statistics by writing
5724 to the "stats" file.
5729 Description: Client mtime is not the same on different clients after utimes
5730 Details : In some cases, the client was using the utimes() syscall on
5731 a file cached on another node. The clients now validate the
5732 ctime from the MDS + OSTs to determine which one is right.
5737 Description: Inability to activate failout mode
5738 Details : lconf script incorrectly assumed that in python string's numeric
5739 value is used in comparisons.
5742 Frequency : always with multiple stripes per file
5744 Description: Inefficient object allocation for mutli-stripe files
5745 Details : When selecting which OSTs to stripe files over, for files with
5746 a stripe count that divides evenly into the number of OSTs,
5747 the MDS is always picking the same starting OST for each file.
5748 Return the OST selection heuristic to the original design.
5753 Description: mount failures may take full timeout to return an error
5754 Details : Under some heavy load conditions it is possible that a
5755 failed mount can wait for the full obd_timeout interval,
5756 possibly several minutes, before reporting an error.
5757 Instead return an error as soon as the status is known.
5759 ------------------------------------------------------------------------------
5761 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5763 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5764 INTEROPERATE with older versions automatically. Please read the
5765 user documentation before upgrading any part of a live system.
5766 * WARNING: Lustre networking configuration changes are required with
5767 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5770 * Support for kernels:
5771 2.6.9-22.0.2.EL (RHEL 4)
5772 2.6.5-7.244 (SLES 9)
5773 2.6.12.6 vanilla (kernel.org)
5776 Severity : enhancement
5777 Bugzilla : 7981/8208
5778 Description: Introduced Lustre Networking (LNET)
5779 Details : LNET is new networking infrastructure for Lustre, it includes
5780 a reorganized network configuration mode (see the user
5781 documentation for full details) as well as support for routing
5782 between different network fabrics. Lustre Networking Devices
5783 (LNDs) for the supported network fabrics have also been
5784 created for this new infrastructure.
5786 Severity : enhancement
5787 Description: Introduced Access control lists
5788 Details : clients can set ACLs on files and directories in order to have
5789 more fine-grained permissions than the standard Unix UGO+RWX.
5790 The MDS must be started with the "-o acl" mount option.
5792 Severity : enhancement
5793 Description: Introduced filesystem quotas
5794 Details : Administrators may now establish per-user quotas on the
5797 Severity : enhancement
5799 Description: Configuration change for the XT3
5800 The PTLLND is now used to run Lustre over Portals on the XT3
5801 The configure option(s) --with-cray-portals are no longer used.
5802 Rather --with-portals=<path-to-portals-includes> is used to
5803 enable building on the XT3. In addition to enable XT3 specific
5804 features the option --enable-cray-xt3 must be used.
5809 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5810 Details : A race between allocating a new llcd and re-getting the llcd_lock
5811 allowed another thread to grab newly-allocated llcd.
5813 Severity : enhancement
5815 Description: 2.6 OST async journal commit and locking fix to improve performance
5816 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5817 2.6 kernels are now async as they already were in 2.4 kernels so
5818 that they can commit concurrently with the network bulk transfer.
5819 For block-allocated files the filter allocation semaphore is held
5820 to avoid filesystem fragmentation during allocation. BKL lock
5821 removed for 2.6 xattr operations where it is no longer needed.
5826 Description: lconf incorrectly determined whether two IP networks could talk
5827 Details : In some more complicated routing and multiple-network
5828 configurations, lconf will avoid trying to make a network
5829 connection to a disjoint part of the IP space. It was doing the
5830 math incorrectly for one set of cases.
5835 Description: Fix for potential infinite loop processing records in an llog.
5836 Details : If an llog record is corrupted/zeroed, it is possible to loop
5837 forever in llog_process(). Validate the llog record length
5838 and skip the remainder of the block on error.
5841 Frequency : occasional (liblustre only)
5843 Description: liblustre could not open files whose last component is a symlink
5844 Details : sysio_path_walk() would incorrectly pass the open intent to
5845 intermediate path components.
5848 Frequency : rare (liblustre only with non-standard tuning)
5849 Bugzilla : 7201 (7350)
5850 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5851 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5852 after it has been released, only do this for referenced locks
5854 Severity : enhancement
5856 Description: specifying an (invalid) directory default stripe_size of -1
5857 would reset the directory default striping
5858 Details : stripe_size -1 was used internally to signal directory stripe
5859 removal, now use "all default" to signal dir stripe removal
5860 as a directory striping of "all default" is not useful
5863 Frequency : common for large clusters running liblustre clients
5865 Description: doing an ls when liblustre clients are running is slow
5866 Details : sending a glimpse AST to a liblustre client waits for every AST
5867 to time out, as liblustre clients will not respond. Since they
5868 cannot cache data we refresh the OST lock LVB from disk instead.
5870 Severity : enhancement
5872 Description: doing an ls at the same time as file IO can be slow
5873 Details : enqueue and other "small" requests can be blocked behind many
5874 large IO requests. Create a new OST IO portal for non-IO
5875 requests so they can be processed faster.
5878 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5880 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5881 created file with mode 000
5882 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5883 allow an ftruncate() as a non-root user to a file with mode 000.
5884 Lustre now respects this flag to disable mode checks when
5885 truncating a file owned by the user
5888 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5891 Description: Revoking locks from clients that went dead or catatonic might take
5893 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5894 cancellation of such locks instant on servers without waiting for
5895 any reply from clients. Clients drops these locks when cancel
5896 notification from server is received without replying.
5899 Frequency : liblustre-only, when liblustre client dies or becomes busy
5901 Description: Doing ls on Linux clients can take a long time with active
5903 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5904 granting such locks to it in the first place if possible. Locks
5905 are taken by proxy on the OST during the read or write and
5906 dropped immediately afterward. Add connect flags handling, do
5907 not grant locks to liblustre clients for glimpse ASTs.
5909 Severity : enhancement
5911 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5912 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5913 sequential reads and behaves badly with applications doing random
5914 reads. Improve it by reading ahead at least read region, and
5915 avoiding excessive large RPC for small reads.
5917 Severity : enhancement
5919 Description: Creating more than 1000 files for a single job may cause a load
5920 imbalance on the OSTs if there are also a large number of OSTs.
5921 Details : qos_prep_create() uses an OST index reseed value that is an
5922 even multiple of the number of available OSTs so that if the
5923 reseed happens in the middle of the object allocation it will
5924 still utilize the OSTs as uniformly as possible.
5929 Description: OST or MDS may oops in ping_evictor_main()
5930 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5931 but doesn't restart at beginning of obd_exports_timed list
5934 Severity : enhancement
5936 Description: improve by-nid export eviction on the MDS and OST
5937 Details : allow multiple exports with the same NID to be evicted at one
5938 time without re-searching the exports list.
5941 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5943 Description: MDS may oops in groups_free()
5944 Details : in rare race conditions a newly allocated group_info struct is
5945 freed again, and this can be NULL. The 2.4 compatibility code
5946 for groups_free() checked for a NULL pointer, but 2.6 did not.
5949 Frequency : common for liblustre clients doing little filesystem IO
5950 Bugzilla : 9352, 7313
5951 Description: server may evict liblustre clients accessing contended locks
5952 Details : if a client is granted a lock or receives a completion AST
5953 with a blocking AST already set it would not reply to the AST
5954 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5957 Frequency : lfs setstripe, only systems with more than 160 OSTs
5959 Description: unable to set striping with a starting offset beyond OST 160
5960 Details : llapi_create_file() incorrectly limited the starting stripe
5961 index to the maximum single-file stripe count.
5964 Frequency : LDAP users only
5966 Description: lconf did not handle in-kernel recovery with LDAP properly
5967 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5969 Severity : enhancement
5971 Description: bind OST threads to NUMA nodes to improve performance
5972 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5973 node and do their allocations there to localize memory access
5975 Severity : enhancement
5977 Description: llmount can determine client NID directly from Myrinet (GM)
5978 Details : the client NID code from gmnalnid was moved directly into
5979 llmount, removing the need to use this or specifying the
5980 client NID explicitly when mounting GM clients with zeroconf
5983 Frequency : if client is started with down MDS
5985 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5986 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5987 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5988 ptlrpc_import_delayed_req() because we didn't check if the
5989 request was interrupted, and we also didn't break out of the
5990 event loop if there was no timeout
5995 Description: data loss during non-page-aligned writes to a single file from
5996 both multiple nodes and multiple threads on one node at same time
5997 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5998 inconsistency led to false short-reads, that were cached and later
5999 used by ->prepare_write() to fill in partially written page,
6000 leading to data loss.
6003 Frequency : always, if lconf --abort_recovery used
6005 Description: lconf --abort_recovery fails with 'Operation not supported'
6006 Details : lconf was attempting to abort recovery on the MDT device and not
6009 Severity : enhancement
6011 Description: remove cleanup logs
6012 Details : replace lconf-generated cleanup logs with lustre internal
6013 cleanup routines. Eliminates the need for client-cleanup and
6016 Severity : enhancement
6018 Description: add support for EAs (user and system) on lustre filesystems
6019 Details : it is now possible to store extended attributes in the Lustre
6020 client filesystem, and with the user_xattr mount option it
6021 is possible to allow users to store EAs on their files also
6023 Severity : enhancement
6025 Description: Add possibility (config option) to show minimal available OST free
6027 Details : When compiled with --enable-mindf configure option, statfs(2)
6028 (and so, df) will return least minimal free space available from
6029 all OSTs as amount of free space on FS, instead of summary of
6030 free spaces of all OSTs.
6032 Severity : enhancement
6034 Description: do not expand extent locks acquired on OST-side
6035 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
6036 by server: they are not cached anyway.
6039 Frequency : when mmap is used/binaries executed from Lustre
6041 Description: Unmmap pages before throwing them away from read cache.
6042 Details : llap_shrink cache now attempts to unmap pages before discarding
6043 them (if unmapping failed - do not discard). SLES9 kernel has
6044 extra checks that trigger if this unmapping is not done first.
6049 Description: lconf didn't resolve symlinks before checking to see whether a
6050 given mountpoint was already in use
6053 Frequency : when migrating failover services
6054 Bugzilla : 6395, 9514
6055 Description: When migrating a subset of services from a node (e.g. failback
6056 from a failover service node) the remaining services would
6057 time out and evict clients.
6058 Details : lconf --force (implied by --failover) sets the global obd_timeout
6059 to 5 seconds in order to quickly disconnect, but this caused
6060 other RPCs to time out too quickly. Do not change the global
6061 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
6063 Severity : enhancement
6064 Frequency : if MDS is started with down OST
6065 Bugzilla : 9439,5706
6066 Description: Allow startup/shutdown of an MDS without depending on the
6067 availability of the OSTs.
6068 Details : Asynchronously call mds_lov_synchronize during MDS startup.
6069 Add appropriate locking and lov-osc refcounts for safe
6070 cleaning. Add osc abort_inflight calls in case the
6071 synchronize never started.
6074 Frequency : occasional (Cray XT3 only)
6076 Description: root not authorized to access files in CRAY_PORTALS environment
6077 Details : The client process capabilities were not honoured on the MDS in
6078 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
6079 been accessed by an authorized user then root was able to access
6080 the file on the local client also. The root user capabilities
6081 are now allowed on the MDS, as this environment has secure UID.
6084 Frequency : occasional
6086 Description: ldiskfs "too long searching" message happens too often
6087 Details : A debugging message (otherwise harmless) prints too often on
6088 the OST console. This has been reduced to only happen when
6089 there are fragmentation problems on the filesystem.
6094 Description: Division by zero in statfs when all OSCs are inactive
6095 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
6096 lov_statfs divides by value returned by lov_get_stripecnt().
6100 Bugzilla : 9489, 3273
6101 Description: First write from each client to each OST was only 4kB in size,
6102 to initialize client writeback cache, which caused sub-optimal
6103 RPCs and poor layout on disk for the first writen file.
6104 Details : Clients now request an initial cache grant at (re)connect time
6105 and so that they can start streaming writes to the cache right
6106 away and always do full-sized RPCs if there is enough data.
6107 If the OST is rebooted the client also re-establishes its grant
6108 so that client cached writes will be honoured under the grant.
6113 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
6114 Details : Now I/O RPCs go to different portal number and (presumably) fast
6115 lock requests (and glimses) and other RPCs get their own service
6116 threads pool that should be able to service those RPCs
6119 Severity : enhancement
6121 Description: Ability to exchange lustre version between client and servers and
6122 issue warnings at client side if client is too old. Also for
6123 liblustre clients there is ability to refuse connection of too old
6125 Details : New 'version' field is added to connect data structure that is
6126 filled with version info. That info is later checked by server and
6130 Frequency : rare, liblustre only.
6131 Bugzilla : 9296, 9581
6132 Description: Two simultaneous writes from liblustre at offset within same page
6133 might proceed at the same time overwriting eachother with stale
6135 Details : I/O lock withing llu_file_prwv was released too early, before data
6136 actually was hitting the wire. Extended lock-holding time until
6137 server acknowledges receiving data.
6140 Frequency : extremely rare. Never observed in practice.
6142 Description: avoid generating lustre_handle cookie of 0.
6143 Details : class_handle_hash() generates handle cookies by incrementing
6144 global counter, and can hit 0 occasionaly (this is unlikely, but
6145 not impossible, because initial value of cookie counter is
6146 selected randonly). Value of 0 is used as a sentinel meaning
6147 "unassigned handle" --- avoid it. Also coalesce two critical
6148 sections in this function into one.
6150 Severity : enhancement
6152 Description: allow liblustre clients to delegate truncate locking to OST
6153 Details : To avoid overhead of locking, liblustre client instructs OST to
6154 take extent lock in ost_punch() on client's behalf. New connection
6155 flag is added to handle backward compatibility.
6157 Severity : enhancement
6158 Bugzilla : 4928, 7341, 9758
6159 Description: allow number of OST service threads to be specified
6160 Details : a module parameter allows the number of OST service threads
6161 to be specified via "options ost ost_num_threads={N}" in the
6162 OSS's /etc/modules.conf or /etc/modprobe.conf.
6166 Bugzilla : 6146, 9635, 9895
6167 Description: servers crash with bad pointer in target_handle_connect()
6168 Details : In rare cases when a client is reconnecting it was possible that
6169 the connection request was the last reference for that export.
6170 We would temporarily drop the export reference and get a new
6171 one, but this may have been the last reference and the export
6172 was just destroyed. Get new reference before dropping old one.
6174 Severity : enhancement
6175 Frequency : if client is started with failover MDS
6177 Description: Allow multiple MDS hostnames in the mount command
6178 Details : Try to read the configuration from all specified MDS
6179 hostnames during a client mount in case the "primary"
6182 Severity : enhancement
6184 Description: Stop sending data to evicted clients as soon as possible.
6185 Details : Check if the client we are about to send or are sending data to
6186 was evicted already. (Check is done every second of waiting,
6187 for which l_wait_event interface was extended to allow checking
6188 of exit condition at specified intervals).
6191 Frequency : rare, normally only when NFS exporting is done from client
6193 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
6195 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
6198 Frequency : always, if srand() is called before liblustre initialization
6200 Description: Liblustre uses system PRNG disturbing its usage by user application
6201 Details : Introduce internal to lustre fast and high-quality PRNG for
6202 lustre usage and make liblustre and some other places in generic
6203 lustre code to use it.
6205 Severity : enhancement
6206 Bugzilla : 9477, 9557, 9870
6207 Description: Verify that the MDS configuration logs are updated when xml is
6208 Details : Check if the .xml configuration logs are newer than the config
6209 logs stored on the MDS and report an error if this is the case.
6210 Request --write-conf, or allow starting with --old_conf.
6212 Severity : enhancement
6214 Description: Handle symlinks in the path when checking if Lustre is mounted.
6215 Details : Resolve intermediate symlinks when checking if a client has
6216 mounted a filesystem to avoid duplicate client mounts.
6221 Description: lconf can hit an error exception but still return success.
6222 Details : The lconf command catches the Command error exception at the top
6223 level script context and will exit with the associated exit
6224 status, but doesn't ensure that this exit status is non-zero.
6229 Description: failure of ptlrpc thread startup can cause oops
6230 Details : Starting a ptlrpc service thread can fail if there are a large
6231 number of threads or the server memory is very fragmented.
6232 Handle this without oopsing.
6235 Frequency : always, only if liblustre and non-default acceptor port was used
6237 Description: liblustre cannot connect to servers with non-default acceptor port
6238 Details : tcpnal_set_default_params() was not called and was therefore
6239 ignoring the environment varaible TCPNAL_PORT, as well as other
6240 TCPNAL_ environment variables
6245 Description: two objects could be created on the same OST for a single file
6246 Details : If an OST is down, in some cases it was possible to create two
6247 objects on a single OST for a single file. No problems other
6248 than potential performance impact and spurious error messages.
6252 Bugzilla : 5681, 9562
6253 Description: Client may oops in ll_unhash_aliases
6254 Details : Client dcache may become inconsistent in race condition.
6255 In some cases "getcwd" can fail if the current directory is
6261 Description: Inode refcounting problems in NFS export code
6262 Details : link_raw functions used to call d_instantiate without obtaining
6263 extra inode reference first.
6267 Bugzilla : 9942, 9903
6268 Description: Referencing freed requests leading to crash, memleaks with NFS.
6269 Details : We used to require that call to ll_revalidate_it was always
6270 followed by ll_lookup_it. Also with revalidate_special() it is
6271 possible to call ll_revalidate_it() twice for the same dentry
6272 even if first occurence returned success. This fix changes semantic
6273 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6274 reference on a request referred from the intent.
6275 ll_intent_release() then releases such a request.
6278 Frequency : rare, normally benchmark loads only
6280 Description: unlinked inodes were kept in memory on the client
6281 Details : If a client is repeatedly creating and unlinking files it
6282 can accumulate a lot of stale inodes in the inode slab cache.
6283 If there is no other client load running this can cause the
6284 client node to run out of memory. Instead flush old inodes
6285 from client cache that have the same inode number as a new inode.
6288 Frequency : SLES9 2.6.5 kernel and long filenames only
6289 Bugzilla : 9969, 10379
6290 Description: utime reports stale NFS file handle
6291 Details : SLES9 uses out-of-dentry names in some cases, which confused
6292 the lustre dentry revalidation. Change it to always use the
6296 Frequency : rare, unless heavy write-truncate concurrency is continuous
6297 Bugzilla : 4180, 6984, 7171, 9963, 9331
6298 Description: OST becomes very slow and/or deadlocked during object unlink
6299 Details : filter_destroy() was holding onto the parent directory lock
6300 while truncating+unlinking objects. For very large objects this
6301 may block other threads for a long time and slow overall OST
6302 responsiveness. It may also be possible to get a lock ordering
6303 deadlock in this case, or run out of journal credits because of
6304 the combined truncate+unlink. Solution is to do object truncate
6305 first in one transaction without parent lock, and then do the
6306 final unlink in a new transaction with the parent lock. This
6307 reduces the lock hold time dramatically.
6310 Frequency : rare, 2.4 kernels only
6312 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6313 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6314 before unmounting their filesystem in order to avoid deadlock.
6315 The kernel_locked() function in 2.4 kernels only checks whether
6316 the kernel lock is held, not whether it is this process that is
6317 holding it as 2.6 kernels do.
6322 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6323 Details : The client ptlrpc code may be trying to reconnect to a down
6324 server before a previous connection attempt has timed out.
6325 Increase the reconnect interval to be longer than the connection
6326 timeout interval to avoid sending duplicate connections to
6330 Frequency : echo_client brw_test command
6332 Description: fix echo_client to work with OST preallocated code
6333 Details : OST preallocation code (5137) didn't take echo_client IO path
6334 into account: echo_client calls filter methods outside of any
6335 OST thread and, hence, there is no per-thread preallocated
6336 pages and buffers to use. Solution: hijack pga pages for IO. As
6337 a byproduct, this avoids unnecessary data copying.
6341 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6342 Description: Client can oops in mdc_commit_close() after open replay
6343 Details : It was possible for the MDS to return an open request with no
6344 transaction number in mds_finish_transno() if the client was
6345 evicted, but without actually returning an error. Clients
6346 would later try to replay that open and may trip an assertion
6347 Simplify the client close codepath, and always return an error
6348 from the MDS in case the open is not successful.
6351 Frequency : rare, 2.6 OSTs only
6353 Description: OST may deadlock under high load on fragmented files
6354 Details : If there was a heavy load and highly-fragmented OST filesystems
6355 it was possible to have all the OST threads deadlock waiting on
6356 allocation of biovecs, because the biovecs were not released
6357 until the entire RPC IO was completed. Instead, release biovecs
6358 as soon as they are complete to ensure forward IO progress.
6360 Severity : enhancement
6362 Description: Support for specifying external journal device at mount
6363 Details : If an OST or MDS device is formatted with an external journal
6364 device, this device major/minor is stored in the ext3 superblock
6365 and may not be valid for failover. Allow detecting and
6366 specifying the external journal at mount time.
6371 Description: Mounting an MDS with pending unlinked files may cause oops
6372 Details : target_finish_recovery() calls mds_postrecov() which returned
6373 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6374 considers this an error and immediately begins cleaning up the
6375 lov, just after starting the mds_lov process
6377 Severity : enhancement
6379 Description: Implement 'lfs df' to report actual free space on per-OST basis
6380 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6381 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6382 usage of objects; '-h' to report in human readable format.
6384 ------------------------------------------------------------------------------
6386 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6393 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6394 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6395 at startup during orphan unlinking. Instead, skip buddy update
6396 before setup, as it will be regenerated after recovery is complete.
6401 Description: If an OST is inactive, its locks might reference stale inodes.
6402 Details : lov_change_cbdata() must iterate over all namespaces, even if
6403 they are inactive to clear inode references from the lock.
6405 Severity : enhancement
6406 Frequency : occasional, if non-standard max_dirty_mb used
6408 Description: Client will block write RPCs if not enough grant
6409 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6410 then the client will block writes while waiting for another RPC
6411 to complete instead of consuming its dirty limit. With change
6412 we get improved performance when max_dirty_mb is small.
6414 Severity : enhancement
6415 Bugzilla : 3389, 6253
6416 Description: Add support for supplementary groups on the MDS.
6417 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6418 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6419 lookups for user supplementary groups into a cache.
6423 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6424 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6425 for liblustre O_EXCL is enforced by MDS.
6429 Description: readdir never returns NULL in liblustre.
6430 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6431 dentry in struct dirent.
6435 Frequency : liblustre only, depends on application IO pattern
6436 Description: liblustre clients evicted if not contacting servers
6437 Details : Don't put liblustre clients into the ping_evictor list, so
6438 they will not be evicted by the pinger ever.
6440 Severity : enhancement
6442 Description: Add ability to evict clients by NID from MDS.
6443 Details : By echoing "nid:$NID" string into
6444 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6445 $NID would be instantly evicted from this MDS and from all active
6446 OSTs connected to it.
6450 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6451 Details : lookup_it_finish() used to query file size from OSTs that was not
6456 Description: service threads change working directory to that of init
6457 Details : Starting lustre service threads may pin the working directory
6458 of the parent thread, making that filesystem busy. Threads
6459 now change to the working directory of init to avoid this.
6463 Frequency : during shutdown only
6464 Description: shutdown with a failed MDS or OST can cause unmount to hang
6465 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6470 Frequency : 2.6 only
6471 Description: chmod/chown may include an extra supplementary group
6472 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6473 supplementary group and if none is specified this is used.
6476 Bugzilla : 5479 (6816)
6478 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6479 Details : If lookup is in progress on a file that is unlinked we might try
6480 to revalidate the inode and fail in revalidate after lookup is
6481 complete and ll_file_open() enqueues the open again but
6482 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6485 Frequency : always, if lconf --abort_recovery used
6487 Description: lconf --abort_recovery fails with 'Operation not supported'
6488 Details : lconf was attempting to abort recovery on the MDT device and not
6491 ------------------------------------------------------------------------------
6493 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6498 Frequency : rare (only unsupported configurations with a node running as an
6500 Bugzilla : 6514, 5137
6501 Description: Mounting a Lustre file system on a node running as an OST could
6503 Details : OSTs now preallocates memory needed to write out data at
6504 startup, instead of when needed, to avoid having to
6505 allocate memory in possibly low memory situations.
6506 Specifically, if the file system is mounted on on OST,
6507 memory pressure could force it to try to write out data,
6508 which it needed to allocate memory to do. Due to the low
6509 memory, it would be unable to do so and the node would
6510 become unresponsive.
6512 Severity : enhancement
6514 Description: Addition of lconf --service command line option
6515 Details : lconf now accepts a '--service <arg>' option, which is
6516 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6518 Severity : enhancement
6520 Description: Failover mode is now the default for OSTs.
6521 Details : By default, OSTs will now run in failover mode. To return to
6522 the old behaviour, add '--failout' to the lmc line for OSTs.
6524 Severity : enhancement
6526 Description: Health checks are now provided for MDS and OSTs
6527 Details : Additional detailed health check information on MSD and OSTs
6528 is now provided through the procfs health_check value.
6531 Frequency : occasional, depends on IO load
6533 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6534 after numerous create/delete cycles
6535 Details : The ext3 inode allocation policy would not allocate new inodes
6536 very well on the OSTs because there are no new directories
6537 being created. Instead we look for groups with free space if
6538 the parent directories are nearly full.
6543 Description: Network or server problems during mount may cause partially
6544 mounted clients instead of returning an error.
6545 Details : The config llog parsing code may overwrite the error return
6546 code during mount error handling, returning success instead
6552 Description: MDS can fail to allocate large reply buffers
6553 Details : After long uptimes the MDS can fail to allocate large reply
6554 buffers (e.g. zconf client mount config records) due to memory
6555 fragmentation or consumption by the buffer cache. Preallocate
6556 some large reply buffers so that these replies can be sent even
6557 under memory pressure.
6561 Frequency : rare (liblustre)
6562 Description: fsx running with liblustre complained that using truncate() to
6563 extend the file doesn't work. This patch corrects that issue.
6564 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6565 ATTR_SIZE and lsm use in llu_setattr_raw.
6569 Frequency : rare, only 2.6 kernels
6570 Description: Unusual file access patterns on the MDS may result in inode
6571 data being lost in very rare circumstances.
6572 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6573 code in the 2.6 kernel caused the inode and/or EA data not to
6574 be read from disk, causing single-file corruption.
6578 Frequency : rare, only 2.6 filesystems using extents
6579 Description: Heavy concurrent write and delete load may cause data corruption.
6580 Details : It was possible under high-load situations to have an extent
6581 metadata block in the block device cache from a just-unlinked
6582 file overwrite a newly-allocated data block. We now unmap any
6583 metadata buffers that alias just-allocated data blocks.
6587 Frequency : filesystems with default stripe_count larger than 77
6588 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6589 Details : lconf specifies an inode size of 4096 bytes when the default
6590 stripe_count is larger than 77. This conflicts with the default
6591 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6592 this case to avoid pinning too much memory for large EAs.
6594 ------------------------------------------------------------------------------
6596 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6601 Frequency : rare (extremely heavy IO load with hundreds of clients)
6603 Description: Client is evicted, gets IO error writing to file
6604 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6605 caused clients to be evicted for AST timeouts. The fixes in
6606 bug 5192 mean we no longer need to have such short AST timeouts
6607 so ldlm_timeout has been increased.
6610 Frequency : occasional during --force or --failover shutdown under load
6611 Bugzilla : 5949, 4834
6612 Description: Server oops/LBUG if stopped with --force or --failover under load
6613 Details : a collection of import/export refcount and cleanup ordering
6614 issues fixed for safer force cleanup
6617 Frequency : only filesystems larger than 120 OSTs
6618 Bugzilla : 5990, 6223
6619 Description: lfs getstripe would oops on a very large filesystem
6620 Details : lov_getconfig used kfree on vmalloc'd memory
6623 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6624 Bugzilla : 6242, 6243
6625 Description: reading from files that had been truncated to a non-zero size
6626 but never opened returned no data
6627 Details : ll_file_read() reads zeros from no-object files to EOF
6632 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6633 Details : There's little harm in aborting MDS/OSS recovery and letting it
6634 try again, so I removed the LASSERT and return an error instead.
6636 Severity : enhancement
6638 Description: New debugging infrastructure for tracking down data corruption
6639 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6640 (b) cover more of the client-side code path, and (c) try to narrow
6641 down where problems occurred
6645 Bugzilla : 3819, 4364, 4397, 6313
6646 Description: Racing close and eviction MDS could cause assertion in mds_close
6647 Details : It was possible to get multiple mfd references during close and
6648 client eviction, leading to one thread referencing a freed mfd.
6650 Severity: : enhancement
6651 Bugzilla : 3262, 6359
6652 Description: Attempts to reconnect to servers are now more aggressive.
6653 Details : This builds on the enhanced upcall-less recovery that was added
6654 in 1.4.2. When trying to reconnect to servers, clients will
6655 now try each server in the failover group every 10 seconds. By
6656 default, clients would previously try one server every 25 seconds.
6661 Description: After recovery, certain operations trigger a failed
6662 assertion on a client.
6663 Details : Failing over an mds, using lconf -d --failover, while a
6664 client was doing a readdir() call would cause the client to
6665 LBUG after recovery completed and the readdir() was resent.
6667 Severity : enhancement
6669 Description: Default groups are now added by lconf
6670 Details : You can now run lconf --group <servicename> without having to
6671 manually add groups with lmc.
6674 Frequency : occasional
6676 Description: Nodes with an elan id of 0 trigger a failed assertion
6679 Frequency : always when accessing e.g. tty/console device nodes
6681 Description: tty and some other devices nodes cannot be used on lustre
6682 Details : file's private_data field is used by device data and lustre
6683 values in there got lost. New field was added to struct file to
6684 store fs-specific private data.
6687 Frequency : when exporting Lustre via NFS
6689 Description: NFSD failed occasionally when looking up a path component
6690 Details : NFSD is looking up ".." which was broken in ext3 directories
6691 that had grown large enough to become hashed.
6694 Frequency : Clusters with multiple interfaces not on the same subnet
6696 Description: Nodes will repeatedly try to reconnect to an interface which it
6697 cannot reach and report an error to the log.
6698 Details : Extra peer list entries will be created by lconf with some peers
6699 unreachable. lconf now validates the peer before adding it.
6702 Frequency : Only if a default stripe is set on the filesystem root.
6704 Description: Setting a default stripe on the filesystem root prevented the
6705 filesystem from being remounted.
6706 Details : The client was sending extra request flags in the root getattr
6707 request and did not allocate a reply buffer for the dir EA.
6710 Frequency : occasional, higher if lots of files are accessed by one client
6711 Bugzilla : 6159, 6097
6712 Description: Client trips assertion regarding lsm mismatch/magic
6713 Details : While revalidating inodes the VFS looks up inodes with ifind()
6714 and in rare cases can find an inode that is being freed.
6715 The ll_test_inode() code will free the lsm during ifind()
6716 when it finds an existing inode and then the VFS later attaches
6717 this free lsm to a new inode.
6721 Bugzilla : 6422, 7030
6722 Description: MDS deadlock between mkdir and client eviction
6723 Details : Creating a new file via mkdir or mknod (starting a transaction
6724 and getting the ns lock) can deadlock with client eviction
6725 (gets ns lock and trying to finish a synchronous transaction).
6728 Frequency : occasional
6729 Description: While starting a server, the fsfilt_ext3 module could not be
6731 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6732 kernels. Previously, lconf would still use the ext3 name
6733 when trying to load modules. Now, it will correctly use
6734 ext3 on 2.4 and ldiskfs on 2.6.
6736 Severity : enhancement
6737 Description: The default stripe count has been changed to 1
6738 Details : The interpretation of the default stripe count (0, to lfs
6739 or lmc) has been changed to mean striping across a single
6740 OST, rather than all available. For general usage we have
6741 found a stripe count of 1 or 2 works best.
6743 Severity : enhancement
6744 Description: Add support for compiling against Cray portals.
6745 Details : Conditional compiling for some areas that are different
6749 Frequency : occasional
6750 Bugzilla : 6409, 6834
6751 Description: Creating files with an explicit stripe count may lead to
6752 a failed assertion on the MDS
6753 Details : If some OSTs are full or unavailable, creating files may
6754 trigger a failed assertion on the MDS. Now, Lustre will
6755 try to use other servers or return an error to the
6759 Frequency : occasional
6761 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6762 caused lock timeout during readahead (since 1.4.2).
6763 Details : Processes doing readahead might match a lock that hasn't been
6764 granted yet if there are overlapping and conflicting lock
6765 requests. The readahead process waits on ungranted lock
6766 (original lock is CBPENDING), while OST waits for that process
6767 to cancel CBPENDING read lock and eventually evicts client.
6769 Severity : enhancement
6771 Description: Initial enabling of flock support for clients
6772 Details : Implements fcntl advisory locking and file status functions.
6773 This feature is provided as an optional mount flag (default
6774 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6775 locking are implemented yet, and those that are are not guaranteed
6776 to be completely correct in production environments.
6777 mount -t lustre -o [flock|noflock] ...
6780 Frequency : occasional
6782 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6783 assertion in the ext3 JBD (journaling) layer.
6784 Details : The b_committed_data struct is protected by the big kernel lock
6785 in 2.4 kernels, serializing journal_commit_transaction() and
6786 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6787 access to this struct. In 2.6 kernels there is finer grained
6788 locking to improve SMP performance of the JBD layer.
6792 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6794 -----------------------------------------------------------------------------
6796 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6798 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6799 versions of Lustre. Please update all servers and clients to
6800 version 1.4.2 or later at the same time. You must also run
6801 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6802 to update the configuration logs.
6804 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6805 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6806 - swab llog records of type '0' so we get proper header size/idx (5861)
6807 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6808 - fix rename of one directory over another leaking an inode (5953)
6809 - avoid SetPageDirty on 2.6 (5981)
6810 - don't re-add just-being-destroyed locks to the waiting list (5653)
6811 - when creating new directories, inherit the parent's custom
6812 striping settings if present parent (3048)
6813 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6814 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6815 - don't hold client locks on temporary worklist from l_lru (5666)
6816 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6817 - automatically evict dead clients (5921)
6818 - Update file size properly in create+truncate+fstat case (6196)
6819 - Do not unhash mountpoint dentries, do not allow removal of
6821 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6822 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6823 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6824 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6825 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6826 - Correctly handle reads of files with no objects (6243)
6827 - lctl recover will also mark a device active if deactivate used (5933)
6829 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6830 - allow --write-conf on an MDS with different nettype than client (5619)
6831 - don't write config llogs to MDS for mounts not from that MDS (5617)
6832 - lconf should create multiple TCP connections from a client (5201)
6833 - init scripts are now turned off by default; run chkconfig --on
6834 lustre and chkconfig --on lustrefs to use them
6835 - upcalls are no longer needed for clients to recover to failover
6837 - add --abort-recovery option to lconf to abort recovery on device
6839 - add support for an arbitrary number of OSTs (3026)
6840 - Quota support protocol changes.
6841 - forward compatibility changes to wire structs (6007)
6842 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6843 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6844 - improved "lustre status" script
6845 - initialize blocksize for non-regular files (6062)
6846 - added --disable-server and --disable-client configure options (5782)
6847 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6848 - Vanilla 2.4.29 support
6849 - increase maximum number of obd devices to 520 (6242)
6850 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6851 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6852 SLES 9 2.6.5 kernels
6853 - limit stripes per file to 160 (the maximum EA size) (6093)
6855 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6858 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6859 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6860 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6861 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6862 - small ext3 extents cleanups and fixes (5733)
6863 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6864 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6865 - avoid needless client->OST connect, fix handle mismatch (5317)
6866 - fix DLM error path that led to out-of-sync client, long delays (5779)
6867 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6868 - fix several locking issues related to i_size (5492,5624,5654,5672)
6869 - don't move pending lock onto export if it is already evicted (5683)
6870 - fix kernel oops when creating .foo in unlinked directory (5548)
6871 - fix deadlock in obdfilter statistics vs. object create (5811)
6872 - use time_{before,after} to avoid timer jiffies wrap (5882)
6873 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6874 - Do not leak request if server was not able to process it (5154)
6875 - If mds_open unable to find parent dir, make that negative lookup(5154)
6876 - don't create new directories with extent-mapping (5909, 5936)
6878 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6879 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6880 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6881 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6882 - support for RHEL4 kernel on i686 (5773)
6883 - provide error messages when incompatible logs are encountered (5898)
6885 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6886 * version 1.4.0.10 (1.4.1 release candidate 1)
6888 - don't keep a lock reference when lock is not granted (4238)
6889 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6890 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6891 - drop import inflight refcount on signal_completed_replay error (5255)
6892 - unlock page after async write error during send (3677)
6893 - handle missing objects in filter_preprw_read properly (5265)
6894 - no transno return for symlink open, don't save no-trasno open (3440)
6895 - don't try to complete elan receive that already failed (4012)
6896 - free RPC server reply state on error (5406)
6897 - clean up thread from ptlrpc_start_thread() on error (5160)
6898 - readahead could read extra page into cache that wasn't ejected (5388)
6899 - prevent races in class_attach/setup/cleanup/detach (5260)
6900 - don't dereference de->d_inode after l_dput of de (5458)
6901 - use "int" for stripe value returned from lock_to_stripe (5544)
6902 - mballoc allocation and error-checking fixes in 2.6 (5504)
6903 - block device patches to fix I/O request sizes in 2.6 (5482)
6904 - look up hostnames for IB nals (5602)
6905 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6906 - don't start multiple acceptors for the same port (5277)
6907 - fix incorrect LASSERT in mds_getattr_name (5635)
6908 - export a proc file for general "ping" checking (5628)
6909 - fix "lfs check" to not block when the MDS is down (5628)
6911 - service request history (4965)
6912 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6913 - create an "evict_client" /proc entry on OSTs, like the MDS has
6914 - fix mount usage message, return errors per mount(8) (5168)
6915 - change grep [] to grep "[]" in tests so they work in more UMLs
6916 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6917 - remove ancient LOV_MAGIC_V0 EA support (5047)
6918 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6919 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6920 - allow readahead allocations to fail when low on memory (5383)
6921 - mmap locking landed again, after considerable improvement (2828)
6922 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6924 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6927 - send OST transaction number in read/write reply to free req (4966)
6928 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6929 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6930 - the watchdog thread now runs as interruptible (5246)
6931 - flock/lockf fixes (but it's still disabled, pending 5135)
6932 - don't use EXT3 constants in llite code (5094)
6933 - memory shortage at startup could cause assertion (5176)
6935 - reorganization of lov code
6936 - single portals codebase
6938 - add extents/mballoc support (5025)
6939 - direct I/O reads in the obdfilter (4048)
6940 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6942 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6945 - send OST transaction number in read/write reply to free req (4966)
6946 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6947 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6948 - the watchdog thread now runs as interruptible (5246)
6949 - handle missing objects in filter_preprw_read properly (5265)
6950 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6951 - drop import inflight refcount on signal_completed_replay error (5255)
6952 - unlock page after async write error during send (3677)
6953 - return original error code on reconstructed replies (3761)
6954 - no transno return for symlink open, don't save no-trasno open (3440)
6956 - add pid to ldlm debugging output (4922)
6957 - bump the watchdog timeouts -- we can't handle 30sec yet
6958 - extra debugging for orphan dentry/inode bug (5259)
6960 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6963 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6964 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6965 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6966 - fix NULL dereference in /proc/sys/portals/routes (4827)
6967 - allow failed mdc_close() operations to be interrupted (4561)
6968 - stop precreate on OST before MDS would time out on it (4778)
6969 - don't send partial-page writes before EOF from client (4410)
6970 - discard client grant for sub-page writes on large-page clients (4520)
6971 - don't free dentries not owned by NFS code, check generation (4806)
6972 - fix lsm leak if mds_create_objects() fails (4801)
6973 - limit debug_daemon file size, always print CERROR messages (4789)
6974 - use transno after validating reply (3892)
6975 - process timed out requests if import state changes (3754)
6976 - update mtime on OST during writes, return in glimpse (4829)
6977 - add mkfsoptions to LDAP (4679)
6978 - use ->max_readahead method instead of zapping global ra (5039)
6979 - don't interrupt __l_wait_event() during strace
6981 - add software watchdogs to catch hung threads quickly (4941)
6982 - make lustrefs init script start after nfs is mounted
6983 - fix CWARN/ERROR duplication (4930)
6984 - return async write errors to application if possible (2248)
6985 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6986 - print NAL number in %x format (4645)
6987 - update barely-supported suse-2.4.21-171 series (4842)
6988 - support for sles 9 %post scripts
6989 - support for building 2.6 kernel-source packages
6990 - support for sles km_* packages
6992 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6995 - ignore -ENOENT errors in osc_destroy (3639)
6996 - notify osc create thread that OSC is being cleaned up (4600)
6997 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6998 - reconstruct ost_handle() like mds_handle() (4657)
6999 - create a new thread to do import eviction to avoid deadlock (3969)
7000 - let lconf resolve symlinked-to devices (4629)
7001 - don't unlink "objects" from directory with default EA (4554)
7002 - hold socknal file ref over connect in case target is down (4394)
7003 - allow more than 32000 subdirectories in a single directory (3244)
7004 - fix blocks count for O_DIRECT writes (3751)
7005 - OST returns ENOSPC from object create when no space left (4539)
7006 - don't send truncate RPC if file size isn't changing (4410)
7007 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
7008 - bind to privileged port in socknal and tcpnal (3689)
7010 - rate limit CERROR/CWARN console message to avoid overload (4519)
7011 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
7012 - basic mmap support (3918)
7013 - kernel patch series update from b1_4 (4711)
7015 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
7018 - avoid crash during MDS cleanup with OST shut down (2775)
7019 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
7020 - don't use bad inodes on the MDS (3744)
7021 - dynamic object preallocation to improve recovery speed (4236)
7022 - don't hold spinlock over lock dumping or change debug flags (4401)
7023 - don't zero obd_dev when it is force cleaned (3651)
7024 - print grants to console if they go negative (4431)
7025 - "lctl deactivate" will stop automatic recovery attempts (3406)
7026 - look for existing locks in ldlm_handle_enqueue() (3764)
7027 - don't resolve lock handle twice in recovery avoiding race (4401)
7028 - revalidate should check working dir is a directory (4134)
7030 - don't always mark "slow" obdfilter messages as errors (4418)
7032 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
7035 - don't close LustreDB during write_conf until it is done (3860)
7036 - fix typo in lconf for_each_profile (3821)
7037 - allow dumping logs from multiple threads at one time (3820)
7038 - don't allow multiple threads in OSC recovery (3812)
7039 - fix debug_size parameters (3864)
7040 - fix mds_postrecov to initialize import for llog ctxt (3121)
7041 - replace config semaphore with spinlock (3306)
7042 - be sure to send a reply for a CANCEL rpc with bad export (3863)
7043 - don't allow enqueue to complete on a destroyed export (3822)
7044 - down write_lock before checking llog header bitmap (3825)
7045 - recover from lock replay timeout (3764)
7046 - up llog sem before sending rpc (3652)
7047 - reduce ns lock hold times when setting kms (3267)
7048 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
7049 - fix NULL deref and obd_dev leak on setup error (3312)
7050 - replace some LBUG about llog ops with error handling (3841)
7051 - don't match INVALID dentries from d_lookup and spin (3784)
7052 - hold dcache_lock while marking dentries INVALID and hashing (4255)
7053 - fix invalid assertion in ptlrpc_set_wait (3880)
7055 - add libwrap support for the TCP acceptor (3996)
7056 - add /proc/sys/portals/routes for non-root route listing (3994)
7057 - allow setting MDS UUID in .xml (2580)
7058 - print the stack of a process that LBUGs (4228)
7060 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
7063 - don't cleanup request in ll_file_open() on failed MDS open (3430)
7064 - make sure to unset replay flag from failed open requests (3440)
7065 - if default stripe count is 0, use OST count for inode size (3636)
7066 - update parent mtime/ctime on client for create/unlink (2611)
7067 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
7068 - free recovery state on server during a forced cleanup (3571)
7069 - unregister_reply for resent reqs (3063)
7070 - loop back devices mounting and status check on 2.6 (3563)
7071 - fix resource-creation race that can provoke i_size == 0 (3513)
7072 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
7073 - more debugging for page-accounting assertion (3746)
7074 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
7075 - avoid deadlock after precreation failure (3758)
7076 - fix race and lock order deadlock in orphan handling (3450, 3750)
7077 - add validity checks when grabbing inodes from l_ast_data (3599)
7079 - add /proc/.../recovery_status to obdfilter (3428)
7080 - lightweight CDEBUG infrastructure, debug daemon (3668)
7081 - change default OSC RPC parameters to be better on small clusters
7082 - turn off OST read cache for files smaller than 32MB
7083 - install man pages and include them in rpms (3100)
7084 - add new init script for (un)mounting lustre filesystems (2593)
7085 - run chkconfig in %post for init scripts (3701)
7086 - drop scimac NAL (unmaintained)
7088 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
7091 - clean kiobufs before and after use (3485)
7092 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
7093 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
7094 - change init script to fail more clearly if not run as root (1528)
7095 - allow clients to reconnect during replay (1742)
7096 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
7097 - don't do DNS lookups on NIDs too small for IP addresses (3442)
7098 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
7099 - fix cond_resched (3554)
7100 - only evict unfinished clients after recovery (3515)
7101 - allow bulk resend, prevent data loss (3570)
7102 - dynamic ptlrpc request buffer allocation (2102)
7103 - don't allow unlinking open directory if it isn't empty (2904)
7104 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
7105 - remove extraneous obd dereference causing LASSERT failure (3334)
7106 - don't use get_cycles() when creating temp. files on the mds (3156)
7107 - hold i_sem when setting i_size in ll_extent_lock() (3564)
7108 - handle EEXIST for set-stripe, set proper directory name (3336)
7110 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
7111 - fix ksocknal_fmb_callback() error messages (2918)
7113 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
7116 - don't copy lvb into (possibly NULL) reply on error (2983)
7117 - don't deref dentry after dput, don't free lvb on error (2922)
7118 - use the kms to determine writeback rpc length (2947)
7119 - increment oti_logcookies when osc is inactive (2948)
7120 - update client's i_blocks count via lvb messages (2543)
7121 - handle intent open/close of special files properly (1557)
7122 - mount MDS with errors=remount-ro, like obdfilter (2009)
7123 - initialize lock handle to avoid ASSERT on error cleanup (3057)
7124 - don't use cancelling-locks' kms values (2947)
7125 - use highest lock extent for kms, not last one (2925)
7126 - don't dereference ERR_PTR() dentry in error handling path (3107)
7127 - fix thread race in portals_debug_dumplog() (3122)
7128 - create lprocfs device entries at setup instead of at attach (1519)
7129 - common AST error handler, don't evict client on completion race (3145)
7130 - zero nameidata in detach_mnt in 2.6 (3118)
7131 - verify d_inode after revalidate_special is valid in 2.6 (3116)
7132 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
7133 - initialize RPC timeout timer earlier for 2.6 (3219)
7134 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
7135 - print nal/nid for unknown nid (3258)
7136 - additional checks for oscc recovery before doing precreate (3284)
7137 - fix ll_extent_lock() error return code for 64-bit systems (3043)
7138 - don't crash in mdc_close for bad permissions on open (3285)
7139 - zero i_rdev for non-device files (3147)
7140 - clear page->private before handing to FS, better assertion (3119)
7141 - tune the read pipeline (3236)
7142 - fix incorrect decref of invalidated dentry (2350)
7143 - provide read-ahead stats and refine rpc in flight stats (3328)
7144 - don't hold journal transaction open across create RPC (3313)
7145 - update atime on MDS at close time (3265)
7146 - close LDAP connection when recovering to avoid server load (3315)
7147 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
7148 - don't leak open file on MDS after open resend (3325)
7149 - serialize filter_precreate and filter_destroy_precreated (3329)
7150 - loop device shouldn't call sync_dev() for nul device (3092)
7151 - clear page cache after eviction (2766)
7152 - resynchronize MDS->OST in background (2824)
7153 - refuse to mount the same filesystem twice on same mountpoint (3394)
7154 - allow llmount to create routes for mounting behind routers (3320)
7155 - push lock cancellation to blocking thread for glimpse ASTs (3409)
7156 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
7157 - fix rare problem with rename on htree directories (3417)
7159 - allow default OST striping configuration per directory (1414)
7160 - fix compilation for qswnal for 2.6 kernels (3125)
7161 - increase maximum number of MDS request buffers for large systems
7162 - change liblustreapi to be useful for external progs like lfsck (3098)
7163 - increase local configuration timeout for slow disks (3353)
7164 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
7166 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7169 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
7170 - don't overwrite extent policy data in reply if lock was blocked (2901)
7171 - drop filter export grants atomically with removal from device (2663)
7172 - del obd_self_export from work_list in class_disconnect_exports (2908)
7173 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
7174 - swab reply message in mdc_close, other PPC fixes (2464)
7175 - fix destroying of named logs (2325)
7176 - overwrite old logs when running lconf --write_conf (2264)
7177 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
7178 - fix race in target_handle_connect (2898)
7179 - mds_reint_create() should take same inode create lock (2926)
7180 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
7181 - don't close files for self_export to avoid uninitialized obd (2936)
7182 - allow MDS with the same name as client node (2939)
7183 - hold dentry reference for closed log files for unlink (2325)
7184 - reserve space for all logs during transactions (2059)
7185 - don't evict page beyond end of stripe extent (2925)
7186 - don't oops on a deleted current working directory (2399)
7187 - handle hard links to targets without a parent properly (2517)
7188 - don't dereference NULL lock when racing during eviction (2867)
7189 - don't grow lock extents when lots of conflicting locks (2919)
7191 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
7194 - account for cache space usage on clients to avoid data loss (974)
7195 - lfsck support in lustre kernel code (2349)
7196 - reduce journal credits needed for BRW writes (2370)
7197 - orphan handling to avoid losing space on client/server crashes
7198 - ptlrpcd can be blocked, stopping ALL progress (2477)
7199 - use lock value blocks to assist in proper KMS, faster stat (1021)
7200 - takes i_sem instead of DLM locks internally on obdfilter (2720)
7201 - recovery for initial connections (2355)
7202 - fixes for mds_cleanup_orphans (1934)
7203 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
7204 - block all file creations until orphan recovery completes (1901)
7205 - client remove rq_connection from request struct (2423)
7206 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
7207 - recovery timer race (2670)
7208 - mdc_close recovey bug (2532)
7209 - ptlrpc cleanup bug (2710)
7210 - mds timeout on local locks (2588)
7211 - namespace lock held during RPCs (2431)
7212 - handle interrupted sync write properly (2503)
7213 - don't try to handle a message that hasn't been replied to (2699)
7214 - client assert failure during cleanup after abort recovery (2701)
7215 - leak mdc device after failed mount (2712)
7216 - ptlrpc_check_set allows timedout requests to complete (2714)
7217 - wait for inflight reqs when ptlrpcd finishes (2710)
7218 - make sure unregistered services are removed from the srv_list
7219 - reset bulk XID's when resending them (caught by 1138 test)
7220 - unregister_bulk after timeout
7221 - fix lconf error (2694)
7222 - handle write after unfinished setstripe, stripe-only getstripe (2388)
7223 - readahead locks pages, leaves pending causing memory pressure (2673)
7224 - increase OST request buffers to 4096 on large machines (2729)
7225 - fix up permission of existing directories in simple_mkdir (2661)
7226 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
7227 - don't assign transno to errored transactions (2742)
7228 - don't delete objects on OST if given a bogus objid from MDS (2751)
7229 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
7230 - if rq_no_resend, then timeout request after recovery (2432)
7231 - fix MDS llog_logid record size, 64-bit array alignment (2733)
7232 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
7233 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
7234 - MDS recovery shouldn't delete objects that it has given out (2730)
7235 - if enqueue arrives after completion, don't clobber LVB (2819)
7236 - don't unlock pages twice when trigger_group_io returns error (2814)
7237 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
7238 - don't write pages to disk if there was an error (1450)
7239 - don't ping imports that have recovery disabled (2676)
7240 - take buffered bytes into account when balancing socknal conn (2817)
7241 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
7242 - reconnect unlink llog connection after MDS reconnects to OST (2816)
7243 - remove little-endian swabbing of llog records (1987)
7244 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
7245 - retry reposting request buffers if they fail (1191)
7246 - grow extent at grant time to avoid granting a revoked lock (2809)
7247 - lock revoke doesn't evict page if covered by a second lock (2765)
7248 - disable VM readahead to avoid reading outside lock extents (2805)
7250 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
7251 - updated kernel patches for hp-2.4.20 kernel (2681)
7253 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
7256 - fix truncated write corruption (2366)
7257 - fix for failed assertion in iopen_connect_dentry (1792,2517)
7259 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
7260 - ensure len(uuid) < 37 in lmc (1171)
7261 - fix ia64 OOPS in llog_test (2255)
7262 - zero end of page at obdfilter for partial page writes (2648)
7263 - don't leave stale dentries around after renames (bug 2428)
7264 - fix timeouts when evicting a client with a single lock held (2642)
7265 - set deadline for the initial HELLO message to drain (2634)
7266 - print out dotted-quad IP addresses in the socknal (2302)
7268 - additional debugging for MDS client eviction problem (2443)
7269 - fix mkfsoptions support for osts (2603, 2604)
7271 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7274 - add series for the vanilla 2.6.0 kernel
7275 - add series for the vanilla 2.4.24 kernel
7276 - add series for a cray x86/64 UL kernel drop
7277 - fix xattr patches for the vanilla 2.4.19 series
7279 - generate true UUIDs in lmc (1171)
7280 - have portals stack dumping break in UML (2466)
7281 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7282 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7283 - wake callback waiting threads on client eviction (2460)
7284 - Add --ptldebug and --subsystem to lmc (1719)
7285 - update assertion to allow safe interrupt allocation
7286 - set rq_no_resend for cancel requests (2432)
7287 - recalculate ptlrpcd timeout after resend (2494)
7288 - call vfs_rmdir when removing pending directories (2368)
7289 - fix renaming a file to itself (2429)
7290 - lmc creates a default one-stripe lov (2454)
7291 - expand procfs space to handle large clusters (2326)
7292 - increase UML stack to avoid overflow
7293 - update lconf's list of debug and subsystem masks
7294 - fix lfs find --obd (2510)
7295 - /proc tunable for disabling filter read caching (2591)
7296 - stop rpm packages from altering slapd.conf (2301)
7297 - disable nagle in the socknal under 0conf (2578)
7298 - choose mds inode size based on stripe count (2572)
7299 - fix kernel-source rpm problems (2516)
7301 - add --disable-doc to avoid pdf generation (2421)
7302 - update documentation, tests, type-os, comments
7303 - avoid format warnings on ia64
7304 - remove the TOE NAL
7305 - tiny code cleanups by removing unused fields
7307 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7310 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7311 - remove the most blatant lies from BUILDING file (2371)
7312 - change default debug level to reasonable production setting
7313 - reduce client side cache size to reduce cache flush time
7314 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7315 - make TCP zerocopy and pinger support enabled by default (2476)
7316 - sync writes completed after process exits caused crashes (2319)
7317 - maintain correct mount count on the MDS (2356)
7318 - backout 1557, because 2316 wasn't really fixed
7319 - better file I/O statistics gathering in /proc
7320 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7321 - another kernel patch to fix zero-copy TCP function export
7322 - don't take duplicate lock when processing re-sent getattr (2420)
7323 - lctl uses obd_self_export instead of creating new conn (2353)
7324 - MDS/OST recovery case which requires object creation asserted (2425)
7325 - move lfs from /usr/sbin to /usr/bin in packages
7326 - fix race between mds_client_add and mds_client_free (2417)
7327 - use kmalloc instead of slabs in portals (2430)
7328 - don't create duplicate records when a failover MDS is present (2442)
7329 - remove unnecessary mount age check (2332)
7330 - don't remove directory inodes from locks prematurely (2451)
7331 - don't break if MDS service name is the same as hostname (2103)
7332 - fix races in client write RPC generation when cache full (2482)
7334 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7337 - remove now-unused request->rq_obd (278)
7338 - if an allocation fails, print out how much memory we've used (1933)
7339 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7340 - add the "configurable stack size" patch to most series files (1256)
7341 - ability to write large log records, for 100+ OST configs (2306)
7342 - fix NULL deref when filter_prep fails (2314)
7343 - fix operator precedence error in filter_sync
7344 - dynamic allocation of socknal TX descriptors (2315)
7345 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7346 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7347 - ensure that CWARN messages in Portals always get to the syslog
7348 - __init/__exit are not for prototype decls (ldlm_init/exit)
7349 - x86-64 compile warning fixes
7350 - fix gateway LMC keyword conflict (2318)
7351 - fix MDS lock inversions in getattr/reint paths (1844)
7352 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7353 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7354 - DLM race condition prevented some lock evictions (2328)
7355 - ENOMEM detection and retry on socknal sends (2230)
7356 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7357 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7358 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7359 - better error messages when a client is rejected during recovery (1505)
7360 - avoid cancelling locks which were never granted, after failure (2330)
7361 - fix i_sem/journal inversion in mds_client_add (2333)
7362 - fix truncate/getattr lock cycle deadlock (2334)
7363 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7364 - fix two rare exit paths which could leak an l_lock() ref (2321)
7365 - fencepost error in MDS/OST orphan recovery (2226)
7366 - make log record alignment 8 bytes (1988)
7367 - lstripe now fails when requested offset > ost_count (2237)
7368 - ensure that all kernel series have a complete list.h (1607)
7369 - fix crashes in special-file operations (2316)
7370 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7372 - allow configurable automake binary, for testing new versions
7373 - small update to the lfs documentation
7375 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7377 * fix negative export reference count in fsfilt_sync (2312)
7379 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7380 * release candidate 0.9.1
7382 - orphans are moved into the PENDING directory for possible recovery
7383 - replayed opens now open by fid for orphan/rename safety (1042)
7384 - last close of an orphan inode generates a transno (683)
7385 - chdir() and mount() now pin the directory entry (1020)
7386 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7387 - discard very old requests without processing them (1502)
7388 - remove some common, well-understood CERRORs (1505)
7389 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7390 - clear "grant" flags in OST replies until OST grant code lands (1644)
7391 - fix read performance by not clobbering i_blksize on client (1598)
7392 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7393 - make LRU size a /proc tunable, clears locks when reduced (707)
7394 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7395 - server locks take references on exports now (1558)
7396 - build fixes for 2.4.20-rh trees (1663)
7397 - return an error from lov_create if all OSCs are inactive (1751)
7398 - fix import levels when a reconnect happens without a timeout (1597)
7399 - exit early from mds_open if we get a lookup error (1749)
7400 - partial page read at EOF wouldn't wait for disk before sending (1642)
7401 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7402 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7403 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7404 - write/truncate lock inversion (1639)
7405 - Don't auto-load obdclass, portals modules during cleanup (1495)
7406 - fix timestamps from jumping to "now" (1763)
7407 - extra journal assertions (1648)
7408 - add an extra multiunlink test (1771)
7409 - fix read_record/write_record API (1776)
7410 - fix leak of offset_extent, possible incorrect i_size later (1772)
7411 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7412 - Debugging for the kqswnal_get_idle_tx problems (1820)
7413 - Allow recovery to be attempted multiple times (1536)
7414 - Write out MDS last_rcvd file after it is first created (1600)
7415 - Fix tx_descriptor leak in failed transmit situations (1827)
7416 - ext3 journaling fixes for assertion failure after IO error (1871)
7417 - class_export_put() on freed export after completion AST error (1896)
7418 - Fix revalidate looping in VFS (1322)
7419 - Don't access a freed export during MDS_REINT timeout (1521)
7420 - Add open-unlink recovery support on the MDS (1673,1764)
7421 - Return an error if no MDS data was read from last_rcvd (1946)
7422 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7423 - Don't setup a disk device that doesn't match exported UUID (317)
7424 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7425 - avoid committing NULL handle in force close
7426 - local.sh is now a one-stripe LOV configuration
7427 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7428 - fix ext3 htree duplicate directory entry corruption (1516)
7429 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7430 - update to new LOV EA format (2097)
7431 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7433 2003-06-15 Phil Schwan <phil@clusterfs.com>
7436 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7437 - per-import/export recovery handling (958, 931, 959)
7438 - multiple last-rcvd slots, for serving multiple FSes (949)
7439 - connections are again shared between multiple imp/exports (963, 964)
7440 - "umount -f" would hang if any requests needed to be sent (393, 978)
7441 - avoid pinning large req buffer by copying for queued messages (989)
7442 - add "uuid" to "lctl device" command to help upcalls (991)
7443 - "open" RPCs with transnos would confuse recovery counters (1037)
7444 - do proper endian conversion of all wire messages (288, 340, 891)
7445 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7446 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7447 - fix ptlrpc_connection leak in target_handle_connect (1174)
7448 - fix import refcounting bug in OST and MDS cleanup (1134)
7449 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7450 - fix very unlikely obd_types race condition (501)
7451 - remove osc_open hack for echo_client (1187)
7452 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7453 - a failure in read_inode2 leads to deadlock (1139)
7454 - cancel ack-locks as soon as transaction is committed (1072)
7455 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7456 - make sure to commitrw after any preprw to avoid deadlock (1162)
7457 - failing to execute a file in a lustre FS would lock inode (1203)
7458 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7459 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7460 - obd_uuid2tgt would walk off the end of the list (1255)
7461 - on IA64 the osc would give portals incorrect bulk size (1258)
7462 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7463 - fix lock inversion caused by new llite matching code (1282)
7464 - limit the number of dirty pages on a client to 10MB (1286)
7465 - timed out locks were not being corrected cancelled (1289)
7466 - fix O_DIRECT above 4GB on IA-32 (1292)
7467 * major user-visible changes
7468 - fail out/fail over policy now controlled by the upcall (993)
7470 - add OBD_PING to check server availability and failure (954)
7471 - lustre messages are now sent in sending host order (288, 340, 891)
7472 - add eadatalen to MDS getattr reply (340)
7473 - OST read replies may contain second buffer, with per-page status (593)
7475 2003-03-11 Phil Schwan <phil@clusterfs.com>
7478 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7479 - failed open()s could cause deadlock; fixed (867, 869)
7480 - stop cancelling OST locks when files are closed (481)
7481 - overlapping XID spaces caused network corruption (851, 853)
7482 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7483 - setattr_raw added, to do single-RPC, server-side setattrs
7484 - lmc/lconf syntax change for OST UUIDs
7485 - fix crashy race condition between ptlrpc_free_req and osc_close
7486 - don't use request in mdc_enqueue if we hit a timeout (889)
7487 - don't set the inode i_size for regular files from the MDS (896)
7488 - handle out of order completion AST (842)
7489 - don't LBUG if a lock request times out after receiving AST (913)
7490 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7491 - if a bad lock AST arrives, send an error instead of dropping entirely
7492 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7493 - fix leak in bulk IO when only partially completed (899, 900, 926)
7494 - fix O_DIRECT for ia64 (55)
7495 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7496 - C-z after timeout could hang a process forever; fixed (977)
7498 - client-side I/O cache (678, 924, 929, 941, 970)
7500 - READPAGE and SETATTRs which don't take server-side locks get
7503 2003-02-11 Phil Schwan <phil@clusterfs.com>
7506 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7507 - Fix fsfilt_extN_readpage() to read a full page of directory
7508 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7509 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7510 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7511 - Most intent operations were removed, in favour of a new RPC mode
7512 that does a single RPC to the server and bypasses most of the VFS
7513 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7514 - Aggressively cancel local locks on DLM servers
7515 - mds_reint_unlink sends EA to the client if it's the last nlink.
7516 client uses that EA to unlink OST objects.
7517 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7518 - recursive symlinks were fixed (439)
7519 - fixed NULL deref in DEBUG_REQ
7520 - filter_update_lastobjid no longer calls sync, which annoyed extN
7521 - fixed multi-client small-writes to a single file problem (445)
7522 - fixed mtime updates during file writes (607)
7523 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7524 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7525 - fixed Linux OST size reporting problem (444, 656)
7526 - OST now updates object mtime with writes or setattr (607, 619)
7527 - client verifies file size before zeroing page past EOF (445)
7528 - OST now writes last allocated objid to disk with allocation (108)
7529 - LOV on echo now works (409)
7531 - mds_reint_unlink sends a new buffer, with the EA included. this
7532 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7533 set if a regular file was being unlinked, and it was the last link
7534 - use PtlGet from the target for bulk writes (315)
7535 - OST now updates object mtime with writes or setattr (607, 619)
7536 - LDLM now has a grant-time callback to revalidate locked items, if
7538 - Many MDS operations were reorganized to combat race conditions
7540 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7541 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7543 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7546 - Fully reactivate OST imports after reconnection (512, others)
7547 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7548 - More graceful error handling for truncating on dead OST (515)
7549 - Don't error out unless we're actually accessing dead stripes (474)
7550 - Fix garbage sizes when stripes are missing (410)
7551 - LRU counters were broken, causing constant lock purge (433, 432)
7552 - garbage on read from stripes with failed OSTs (441)
7553 - mark OSCs as active before reconnecting during recovery (438)
7554 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7555 - lfind did not preserve OST order in output (443)
7556 - symlinks cause hung clients, incorrect data (439)
7557 - stop dereferencing request after dropping refcount (457)
7558 - don't LASSERT(spin_is_locked) on non-SMP (455)
7559 - fixes for many rename() bugs
7560 - fstat didn't correctly synchronize attributes (399)
7561 - server must handle lock cancellation during blocking AST prep (487)
7562 - bulk descriptors were free()d too soon (511)
7563 - fix paths in lconf, which would load incorrect modules (451, 507)
7564 - fix confusing lconf 'host not found' error message (386)
7565 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7566 - fix race condition in mdc_blocking_ast() for inode access (526)
7567 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7568 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7569 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7570 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7571 - fix lstripe in several ways (kernel side) (527)
7572 - fix request leak in ldlm_cli_enqueue (262)
7573 - incorrect OSC was marked inactive after OST failure
7574 - call mds_fs_cleanup before unmounting filesystem (524)
7575 - fix races between taking ns_lock and ldlm_lock_change_resource
7576 - fix races updating LOV export open file list
7577 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7578 - fix recovery NULL deref in ldlm_cli_cancel_unused
7579 - fix some DLM races by using new hash table for lock handles (419)
7580 - permit the client to specify desired inodes, at replay
7581 - duplicate requests when we queue them for replay reintegration
7582 - fix last_rcvd offset calculation
7583 - sync after each recovered transaction, so we always make progress
7584 - never, not always, ERESTART requests without transnos
7585 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7587 - skip replay if the MDS says that the client is already connected
7588 - don't check for a recovery-enabled export to match lctl's UUID
7589 - don't INC_USE_COUNT for phantom exports
7590 - don't crash when cleaning up phantom exports (567)
7591 - don't double-finish or set replay data for errored mdc_open requests
7592 - abort requests when they time out, so we don't get old replies
7593 - send/receive replies for AST messages again
7594 - if the client says that it doesn't have the lock, cancel it on the
7596 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7597 mark it as destroyed, it will all work out when decref is called
7598 - fix module use counts (22, 581)
7600 - ASTs now expect a reply (server cancels lock on error reply)
7602 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7605 - fix many simultaneous client startup (392)
7606 - fix dentry->d_it clobbering
7607 - credentials weren't being shipped for readdir/getattr operations
7608 - remove invalid assertions triggered during some concurrent MD
7610 - proper Lustre versions added (336, 389)
7611 - fix memory leak for create error case (398)
7612 - fix LOV locking bug that would get cli/srv out of sync
7613 - fix echo client over LOV (409)
7614 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7615 - fix double-O_EXCL intent crash (424)
7616 - avoid sending multiple lock CANCELs (352)
7618 - MDS can do multi-client recovery (modulo bugs in new code)
7620 - many updates, edits, cleanups
7622 2002-11-18 Phil Schwan <phil@clusterfs.com>
7625 - fix null d_it dereference (346)
7626 - fix full OST/dbench hang (333)
7627 - fix permission problem with file removal (286)
7628 - fix removal of OSCs from LOV when they fail
7629 - fix NULL deref during bulk timeout (214)
7630 - fix problems related to multiple filesystems on one MDS (241)
7631 - fixed serious subtle metadata locking bugs
7632 - free locks on clients when inodes are removed due to memory
7634 - fix inode pointer in lock data (285)
7635 - partial support for multiple MDS on a single host (241)
7636 - data locks weren't cancelled at clear_inode time (290, 311)
7637 - intent locks could lead to unbounded lock growth (205)
7638 - added a maximum lock count, an LRU list, and a flusher
7639 - fix multiple rename (365)
7640 - properly abstracted the echo client
7641 - OSC locked 1 byte too many; fixed
7642 - rewrote brw callback code:
7643 - fixed recovery bugs related to LOVs (306)
7644 - fixed too-many-pages-in-one-write crash (191)
7645 - fixed (again) crash in sync_io_timeout (214)
7646 - probably fixed callback-related race (385)
7648 - Add capability to MDS protocol
7649 - LDLM cancellations and callbacks on different portals
7651 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7654 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7655 - defer open object destruction to close time (601981, 138)
7656 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7657 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7658 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7659 - don't call obd_setattr in ll_file_release for destroyed objects
7660 * protocol change to lustre_msg: move |version| and add |flags|
7661 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7662 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7663 * added replay of create, unlink, link and rename operations during
7664 MDS failover; recovery should be much more robust now
7665 * remove failed OSCs from LOVs (only lov_create uses this so far)
7666 * the lustre-HOWTO was brought (more) up to date (582544)
7668 2002-10-23 Phil Schwan <phil@clusterfs.com>
7671 - in-use dentries weren't being reused properly (617851)
7672 - prevent multiple LDLM setup (599178)
7673 - fix LOV size calculations for truncate (617853)
7674 - fix client handling of MDS intent errors (POSIX)
7675 - fix permission bug in lovstripe.c test (624321)
7676 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7677 - truncate past end of file could corrupt data
7678 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7679 - a race in recovery could return ETIMEDOUT to apps (623947)
7680 - building outside the source directory was fixed
7681 * the lustre-HOWTO was brought (more) up to date (582544)
7682 * major progress was made on recovery functionality
7684 2002-10-10 Phil Schwan <phil@clusterfs.com>
7687 - recovery deadlock fix
7688 - rm -rf causes LBUG fix (617817)
7689 - file open by multiple tasks fix (618962)
7690 - directory permissions bugs (602707 and 620007)
7691 - journal_stop fixed with locking (611313)
7692 - O_APPEND failures resolved (618273, perhaps 614459)
7693 - lconf PATH fix (619770)
7694 - IA64 build fix (621450)
7695 - RPC buffer sizes scale with amount of memory
7697 2002-10-01 Phil Schwan <phil@clusterfs.com>
7700 - locks would be cancelled without throwing away data pages,
7701 resulting in inconsistent data (605627)
7702 - inode attributes were not always being refreshed (605627, 612449)
7703 - lconf now continues to cleanup after lctl reports an error
7704 - MDS now enforces user permissions (602707)
7705 - lprocfs cleanup fixed, but not yet enabled (614157)
7706 - fixed infinite server hang, should a client not respond to an AST
7707 - avoid going into recovery if user calls readlink() with a buffer
7708 that's too small (613941)
7709 - AST RPCs no longer require replies (614867) -- this may be changed
7710 - don't crash server if client sends an IOV that's too big (611336)
7711 - fixed lock conversion deadlock (611892)
7712 - fixed the following of symlinks (614622)
7713 * recovery: the server can remove locks from a client that dies, other
7714 clients can make progress
7715 * more extN patch fixes
7716 * compile-time configurable ptlrpc buffer allocations
7718 - collaborative read cache document
7719 - Lustre Lite Performance CDR document-in-progress
7721 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7724 - fix typo in patch-2.4.18
7726 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7729 - clear ptlrpc request each time in handle_incoming_request()
7730 - unlink of files now destroys the object on the OST
7732 2002-09-19 Peter Braam <braam@clusterfs.com>
7734 * add hard link support
7735 * change obdfile creation method
7736 * kernel patch changed
7738 2002-09-19 Peter Braam <braam@clusterfs.com>
7741 - stack overflow bug in extN fixed
7743 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7745 * documentation updates
7746 - add man pages for config tools
7747 - update tests/README to describe testing with new config tools
7748 - finish metadata API descriptions
7749 * bug fixes and cleanups
7750 - statfs workaround for 16TB limit
7751 - LOV stripe allocation improved, can stripe on subset of OSTs
7752 - LOV file size/IO offset was wrong for files > 4GB in size
7753 - object EA data was being dropped, caused files to be unreadable
7754 - memory overflow with non-LOV OST caused memory corruption
7755 - fixed regression tests to work with new config tools, obdfilter
7756 - fixed bug when directory size became larger than 1 block
7757 - fixed bug (for single client case) when PWD was deleted
7758 - invalidate local directory pages when doing intent-based ops
7759 - avoid LDLM oops when lock callback contained bad data
7761 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7763 * documentation updates
7764 * bug fixes and cleanups
7765 - configuration tools
7768 - 64-bit compile warnings
7769 - 64-bit internal statfs data
7771 * test_brw on persistent OST devices
7773 * lprocfs (disabled)
7775 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7777 * documentation updates
7778 * bug fixes and cleanups
7779 * configuration tools
7781 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7784 * many small fixes to 0_5_4
7785 * io/network handling
7786 * thinkos in MDS operations
7788 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7791 * crucial basic fixes to 0.5.3
7792 * IOR, Iozone work over Elan
7795 2002-08-07 Phil Schwan <phil@clusterfs.com>
7796 * version 0_5_3, our first alpha
7797 * we use the new Portals iovs
7798 * documentation updates
7799 * bug fixes and cleanups
7800 * small changes in the DLM wire protocol
7802 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7803 * version 0_5_1 with some initial stability,
7804 * locking on MD and file I/O.
7805 * documentation updates
7806 * several bug fixes since 0.5.0
7807 * small changes in wire protocol
7809 2002-07-18 Phil Schwan <phil@clusterfs.com>
7811 * delivered as Lustre Light Alpha
7812 * fixed a crash after handling invalid MDS requests
7813 * fixed directory pages for architectures with non-4k pages sizes
7815 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7816 * release version v0_4_4
7817 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7819 * Fixes a number of interruption problems with OST operations.
7820 * Update documentation for portals header changes
7821 * Move all wire protocol structs/defines to lustre_idl.h
7822 * Fixes symlink length bug.
7823 * Add tcpdump to repository.
7825 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7826 * release version v0_4_3
7827 * Fixes statfs for inodes on extN.
7828 * Fixes bug in runtests which would delete /etc/hosts.
7829 * Use 64-bit object IDs wherever possible (not into VFS though)
7830 Remove ost_get_info, which is unused by lustre, and out of date.
7832 2002-07-03 Peter Braam <braam@clusterfs.com>
7833 * release version v0_4_2 Fixes a lookup error (type not passed)
7834 * move forward to head of Portals
7835 * move forward to latest Lustre kernel
7837 2002-06-25 Peter Braam <braam@clusterfs.com>
7838 * release version v0_4_1. Hopefully stable on single node use.